Open Access Atmos. Chem. Phys., 14, 4539–4562, 2014 Atmospheric www.atmos-chem-phys.net/14/4539/2014/ doi:10.5194/acp-14-4539-2014 Chemistry © Author(s) 2014. CC Attribution 3.0 License. and Physics

Borneo vortex and mesoscale convective rainfall

S. Koseki1, T.-Y. Koh2,3, and C.-K. Teo2 1Temasek Laboratories, Nanyang Technological University, Singapore 2School of Physical and Mathematical Science, Nanyang Technological University, Singapore 3Earth Observatory of Singapore, Nanyang Technological University, Singapore

Correspondence to: T.-Y. Koh ([email protected])

Received: 28 May 2013 – Published in Atmos. Chem. Phys. Discuss.: 13 August 2013 Revised: 10 March 2014 – Accepted: 22 March 2014 – Published: 9 May 2014

Abstract. We have investigated how the Borneo vortex de- 1 Introduction velops over the equatorial South China Sea under cold surge conditions in December during the Asian winter monsoon. Composite analysis using reanalysis and satellite data sets Activity of the Asian monsoon is related to the seasonal has revealed that absolute vorticity and water vapour are migration of the Intertropical Convergence Zone (ITCZ; transported by strong cold surges from upstream of the e.g. Hubert et al., 1969) and plays a key role in the climate South China Sea to around the Equator. Rainfall is corre- and weather systems in the Maritime Continent. Basically, spondingly enhanced over the equatorial South China Sea. the Asian winter monsoon is forced by land–sea thermal A semi-idealized experiment reproduced the Borneo vortex contrast between the cool Eurasian continent (high surface over the equatorial South China Sea during a “perpetual” pressure) and the warm sea surface in East and Southeast cold surge. The Borneo vortex is manifested as a meso- Asia (low surface pressure) in boreal winter (e.g. Tomita and α cyclone with a comma-shaped rainband in the northeast Yasunari 1996). Northeasterly winds prevail over the South sector of the cyclone. Vorticity budget analysis showed that China Sea (Fig. 1a) in the lower troposphere. This northeast- the growth/maintenance of the meso-α cyclone was achieved erly monsoonal flow is cool and dry because the air mass mainly by the vortex stretching. This vortex stretching is due originates from the Eurasian landmass and the monsoon en- to the upward motion forced by the latent heat release around hances upward sensible and latent heat fluxes over the South the cyclone centre. The comma-shaped rainband consists of China Sea. The enhanced surface fluxes contribute partly to clusters of meso-β-scale rainfall cells. The intense rainfall the generation of the cold tongue in sea surface temperature in the comma head (comma tail) is generated by the conflu- (SST) (Chen et al., 2003) in the South China Sea during the ence of the warmer and wetter cyclonic easterly flow (cy- winter monsoonal season. clonic southeasterly flow) and the cooler and drier northeast- Occasionally, the low-level northeasterly winds from the erly surge in the northwestern (northeastern) sector of the cy- Eurasian continent associated with the Asian winter mon- clone. Intense upward motion and heavy rainfall resulted due soon can abruptly intensify. Such rapid intensification of to the low-level convergence and the favourable thermody- the Asian winter monsoon which is related to the activ- namic profile at the confluence zone. In particular, the con- ity of the Siberian High (e.g. Joung and Hitchman, 1982; vergence in the northwestern sector is responsible for main- Takaya and Nakamura, 2005) is commonly known as a cold tenance of the meso-α cyclone system. At both meso-α and air outbreak or cold surge (although by the time it reaches meso-β scales, the convergence is ultimately caused by the the Maritime continent, the air temperature has warmed deviatoric strain in the confluence wind pattern but is signif- up considerably due to sensible heat fluxes from the un- icantly self-enhanced by the nonlinear dynamics. derlying South China Sea). In the mid-latitudes, the cold surge forces the Japan Sea Polar Airmass Convergence Zone (JPCZ; e.g. Asai, 1988) and induces cumulus convection over the Sea of Japan which subsequently brings heavy snowfall around coastal northern Japan (Nagata, 1987, 1993; Tsuboki

Published by Copernicus Publications on behalf of the European Geosciences Union. 4540 S. Koseki et al.: Borneo vortex and mesoscale convective rainfall

(a) local sea/land breeze circulation (e.g. Estoque, 1962) and 18N Hainan Luzon through the topography lifting effect (e.g. Seko et al., 2008). Island Island Moreover, cold surges in the Maritime continent often spin 12N up meso-α-scale cyclonic circulation known as Borneo vor- South 6N China Sea tices (e.g. Chen et al., 1986; Lau and Chang, 1987; Johnson Malay Peninsula and Houze, 1987; Juneng and Tangang, 2010; Braesicke et Borneo Eq al., 2012) near the Borneo Island. Chang et al. (2005) stud- Java Sea 6S ied the variability in the Borneo vortex and cold surge us- ing satellite observational and reanalysis data sets. They con- 90E 100E 110E 120E 130E cluded that deep cumulus convection tends to occur fre- (m/s) (b) quently over the South China Sea in the presence of Bor- 20 18 neo vortices. A Borneo vortex generally has a shallow ver- 16 tical structure, but in the extreme case, it could intensify 14 into a tropical storm, and in one instance, even a typhoon 12 10 with a deep vortex tube and a well-formed eyewall despite 8 its proximity to the Equator. This singular event is the ty- 6 phoon Vamei which formed over the equatorial South China 4 Sea (around 1.5◦ N) on 26 December 2001 (e.g. Chang et 2 1981 1982 1983 1984 1985 1986 1987 1988 1989 al., 2003; Chambers and Li, 2007). Chang et al. (2003) in- 20 dicated that the persistence of the cold surge over the South 18 16 China Sea and the quasi-stationary location of the Borneo 14 vortex over the sea for more than 1 week were the necessary 12 pre-conditions for this exceptional case of tropical cycloge- 10 8 nesis. Chambers and Li (2007) ran a model simulation and 6 diagnosed that rapid development in the potential vorticity is 4 responsible for the typhoon generation. 2 1990 1991 1992 1993 1994 1995 1996 1997 1998 While tropical cyclogenesis rarely occurs over the equa- 20 torial South China Sea, the formation of heavily precipitat- 18 16 ing meso-β convective systems in the presence of Borneo 14 vortices is a common occurrence and brings heavy rainfall 12 to the surrounding coastal regions during the boreal winter 10 8 monsoon. There is room for more research to understand the 6 formation and the organization of these smaller but more fre- 4 quent systems. While Chen et al. (2012) showed that the 2 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 interannual variation in rainfall in the region is related to the synoptic-scale cyclonic disturbances over and around the Figure 1. (a) The geography of the Maritime Continent. The black box denotes the area for CS IndexFig.. (b) C 1.ol (a)d SuTherge (C geographyS) Index obta ofined the from Maritime daily JRA2 continent.5/JCDAS in TheDecem blackber fro boxm 1981 to South China Sea, the mechanisms through which mesoscale 2008denotes. The thick the dash areaed lin fore is t Coldhe clim Surgeatology (CS)of CS Index.Index and(b) thinCS dot Indexlines are obtained the climatology convective systems can develop within the disturbances have ± one standard deviation of CS Index. CS Index is defined as area-averaged wind speed at 850 hPa ofromver 7.5 daily°N to 12 JRA25/JCDAS.5°N and 110°E to 1 in15° DecemberE. The lines ar frome disjoi 1981nt betwe toen 2008.years. The not been clarified. Given that there are few diagnostic or dy- thick dashed line is the climatology of the CS Index and thin dotted namical analyses on this topic, we shall explore the mecha- ± lines are the climatology 1 standard deviation of CS Index. CS nism behind the development of the Borneo vortex over the Index is defined as area-averaged wind speed at 850 hPa over 7.5 to equatorial South China Sea and the associated generation and 12.5◦ N and 110 to 115◦ E. The lines are disjoint between years. organization of deep cumulus convection. In this paper, we examine the physical and the dynami- cal connections among different spatial scales: the synoptic and Asai, 2004; Kawashima and Fujiyoshi, 2005). The cold scale for the monsoon cold surge, the meso-α scale for the surge induces a sudden depression in temperature and an in- Borneo vortex and the meso-β scale for cumulus convection. crease in the sea surface pressure in the coastal sea off south- In the first part of the paper, we elucidate through use of re- ern China (e.g. Hong et al., 2009; Lu and Chang, 2009). analysis data and satellite observations, the difference in the In the Southeast Asia region, the cold surge persists for a synoptic-scale anomalies during cold surge events and reveal few days to about 1 week over the South China Sea and the dynamical relation between the cold surge, the Borneo causes severe hazards such as heavy rainfall and flooding vortex and mesoscale convective systems. This gives us an (e.g. Wangwongchai et al., 2005; Tangang et al., 2008; Hat- idea on how the cold surge creates a favourable synoptic- tori et al., 2011; Trilaksono et al., 2012). The northeasterly scale condition both dynamically and thermodynamically for winds can trigger convection through interaction with the

Atmos. Chem. Phys., 14, 4539–4562, 2014 www.atmos-chem-phys.net/14/4539/2014/ S. Koseki et al.: Borneo vortex and mesoscale convective rainfall 4541 the generation of the Borneo vortex heralding cumulus con- NHM simulations of rainfall in the Maritime continent on vection and heavy rainfall in the Maritime continent. weather timescales and showed the consistency of the model The second part of this paper involves a numerical ex- with observations qualitatively. periment using a non-hydrostatic model in order to examine In this study, the horizontal and vertical resolutions of how deep cumulus convection can be formed in the presence NHM are set to be 10 km and 40 vertical layers (the model of a Borneo vortex during a cold surge event, based on the top is at 22 055 m using a terrain-following coordinate with favourable conditions for Borneo vortex formation inferred 34 levels in the troposphere including levels in the lowest from the results of the first part. Such mesoscale details can- first kilometre). We select the Kain–Fritsch cumulus scheme not be captured by the low-resolution of data sets available in (Kain, 2004), the six-class cloud microphysics including wa- this region. Because the Borneo vortex is forced by the cold ter vapour, cloud water, cloud ice, rain, snow and grau- surge mechanically (which is unlike a typical tropical cy- pel (Ikawa and Saito, 1991), the improved Mellor–Yamada clone or typhoon), we analyse the vorticity tendency budget level-3 turbulent closure scheme (Nakanishi, 2001; Nakan- to investigate the mechanism of growth and maintenance of ishi and Niino, 2004, 2006), and the four-layer land surface the Borneo vortex. In addition, our analysis of the divergence scheme (Kumagai, 2004). The topography and land use data tendency budget gives further understanding of the mecha- are obtained from the Global 30 arc second Elevation Data nism, because divergence and vorticity tendencies describe Set (GTOPO30) and the Global Land Cover Characterization together the dynamical evolution of a fluid. Because the Bor- (GLCC) of the United States Geological Survey (USGS). neo vortex is essentially a result of the dynamical forcing Further details of the model configuration such as the bound- by the cold surge, the divergence tendency budget analysis ary and initial conditions of numerical model are given in is able to show some new aspects of the generation and or- Sect. 5. ganization of cumulus convection associated with the Bor- neo vortex at the meso-α scale without detracting from the importance of latent heat release in providing energy, and in 3 Composite analysis of cold surge events maintaining the low-level convergence in the convective cells at the meso-β scale. In this section, we examine the observed atmospheric fields The remainder of the paper is organised as follows: in when strong cold surges are present over the South China Sect. 2, we describe the details of the data sets and model Sea by comparing with the cases when no cold surge events used in this study. We present the results of the data analysis are present using the JRA25/JCDAS reanalyses and TRMM for cold surges and the associated Borneo vortices in Sects. 3 rainfall data. and 4. The results of the numerical experiment are presented in Sect. 5. In Sect. 6, we synthesize and discuss the results 3.1 Synoptic-scale atmospheric circulation of Sects. 3, 4 and 5. Finally, concluding remarks are made in Sect. 7. To classify the strong and no cold surge cases, we first com- pute the Cold Surge (CS) Index (Chang et al., 2005) de- fined as the daily area-averaged wind speed at 850 hPa in 2 Data and numerical model the rectangular domain bounded within 7.5 to 12.5◦ N and 110 to 115◦ E. Figure 1b shows the daily December CS In- To elucidate the synoptic-scale atmospheric features associ- dex from 1981 to 2008 obtained from the JRA25 and JC- ated with cold surges, we use the daily Japanese 25-year Re- DAS reanalyses. We define a CS Index larger (smaller) than analysis (JRA25, Onogi et al., 2007) and the Japanese Mete- the climatology plus (minus) one standard deviation of the orological Agency (JMA) Climate Data Assimilation System CS Index as a strong (no) surge case abbreviated henceforth (JCDAS) data set at 1.25◦ × 1.25◦ horizontal resolution with by SS (NS). SS are present almost every year. They occa- 23 vertical (pressure) levels in December from 1981 to 2008 sionally persist for 1 week or longer, as e.g. in 1984, 1990, (the JRA25 data set is used from 1981 to 2004 while its suc- 1993, 1998, 1999, 2001, 2005, and 2006. It has been sug- cessor, the JCDAS data set is used from 2005 to 2008). In gested that the strong background monsoon wind played an addition, we analyse the 3-hourly rainfall data for the month important role in the formation of tropical typhoon Vamei of December from 1998 to 2008 obtained from the Tropical (Chang et al., 2003) at the record-low latitude of 1.5◦ N Rainfall Measurement Mission (TRMM) version 3B42 data in 2001. In addition, the tropical storm Gil (http://www. set at horizontal resolution of 0.25◦ × 0.25◦ (Huffman et al., weather.gov.hk/publica/tc/tc1998.pdf) occurred after a SS 2007). over the South China Sea between 9 and 12 December 1998 The Non-Hydrostatic Model version.2010-May-10 estimated from the Regional Specialized Meteorological (NHM, Saito et al., 2006, 2007), an operational weather Centre (RSMC) Tokyo Best Track (http:www/jma.go.jp/jma/ prediction model capable of resolving mesoscale dynamics, jma-eng/jma-center/rsmc-hp-pub-eg/RSMC_HP.htm) of the developed by the JMA is used to conduct the numerical JMA. The CS Index for January has been examined from experiment in this study. Hayashi et al. (2008) evaluated the 1981 to 2008 as well (not shown) with similar occurrences www.atmos-chem-phys.net/14/4539/2014/ Atmos. Chem. Phys., 14, 4539–4562, 2014 4542 S. Koseki et al.: Borneo vortex and mesoscale convective rainfall

Fig. 2. The climatology (a, b) and composite anomalies during NS (c, d) and SS (e, f) of horizontal wind (a, c, e) and absolute vorticity (b, d, f) at 850 hPa in December obtained from JRA25/JCDAS. The thick black contour denotes zero value. Positive (negative) values are denoted by solid (dashed) lines in (b), (d) and (f). The grey shading denotes statistical significance at 95 % confidence level. Contour interval is 1.0 × 10−4 s−1 for (b) and 0.2 × 10−4 s−1 for (d) and (f). of SS and NS albeit under different climatological wind northwesterly when they extend into the Java Sea due to the conditions. change in the sign of Coriolis parameter across the Equa- We exclude those days of SS associated with the two ex- tor. During SS (NS), Fig. 2e (Fig. 2c) shows how the winter treme cases of Vamei (20–24 December 2001) and Gil (5– monsoon is reinforced (weakened) over the South China Sea 8 December 1998) from our analysis so as to focus on the and to a lesser extent over the Java Sea. The climatology in typical cases of mesoscale circulation and convection associ- Fig. 2b shows an enhanced positive absolute vorticity extend- ated with strong cold surges without being biased by the rare ing from the west of Luzon Island of the Philippines over extreme events which may involve atypical tropical cycloge- the South China Sea towards the Equator. During SS (NS), netic dynamics. Thus, the number of SS and NS available for Fig. 2f (Fig. 2d) shows positive (negative) difference in ab- our composite study is 133 and 156 days respectively for the solute vorticity over the South China Sea off the northwest- month of December from 1981 to 2008. This translates to ern coast of Borneo and between Sumatra and Borneo, while 15 and 18 % in terms of frequency of occurrences of SS and negative (positive) difference is found around Indochina. The NS events respectively. distribution of the absolute vorticity difference is approxi- In the climatology shown in Fig. 2a, the northeasterly win- mately opposite during NS and SS. The ratio of the differ- ter monsoon winds prevailing over the South China Sea turn ence in absolute vorticity to climatology of absolute vorticity

Atmos. Chem. Phys., 14, 4539–4562, 2014 www.atmos-chem-phys.net/14/4539/2014/ S. Koseki et al.: Borneo vortex and mesoscale convective rainfall 4543 is about 30 % in both the SS and the NS off the northwestern (a) (b) coast of Borneo. 18N 18N

12N 12N 3.2 Synoptic-scale atmospheric transport 6N 6N

We diagnose from the reanalyses the isobaric divergence of Eq Eq absolute vorticity flux defined as ∇p · (V ζa), where V is the 6S 6S horizontal wind vector (u, v), ζa is the cross-isobaric com- 90E 100E 110E 120E 130E ponent of the absolute vorticity, and ∇p is the isobaric gradi- 90E 100E 110E 120E 130E ent operator. The isobaric framework is adopted here because (hPa) (c) (hPa) (d) the solenoid term in the vorticity equation when expressed in 200 200 height coordinate is absent in the pressure coordinate form of the equation. This allows us to simplify the vorticity bud- 400 400 get analysis in Sect. 5. We describe the difference between

SS/NS composite and climatology in this section. 600 600 Strong convergence of absolute vorticity flux is located over the South China Sea east of Indochina and the Malay 800 800 Peninsula during SS while strong divergence of similar mag- nitude is found further north in the west of Luzon Island 1000 1000 (Fig. 3a). This indicates that as the winter monsoon blows 90E 100E 110E 120E 130E 90E 100E 110E 120E 130E from northeast to southwest (Fig. 2a and c), high abso- Fig. 3. (a)–(b) Divergence of absolute vorticity and moist static lute vorticity is transported southwestward from the north- energy flux anomalies (SS composite minus climatology) obtained ern tropical latitudes toward the Equator. In particular, the from JRA25/JCDAS at 850 hPa in December. (c) and (d) Pressure– 1 Figure 3. (a)-(b) Divergence of absolute vorticity and moist static energy fluxes anomalies (SS◦ anomalous isobaric convergence of absolute vorticity flux in longitude2 composite sectionsminus climatology) of (a) obtainedand (b)from respectivelyJRA25/JCDAS at 850 averaged hPa in December. between (c) and 0 (d) 3 Pressure-longitude◦ sections of (a) and (b) respectively averaged between 0° and 4°N. Black solid the equatorial South China Sea between the Malay Penin- and4 (dashed) 4 N. line Black is positive solid (negative) (dashed) value. Contour line interval is positive is 5×10-5 s-1 (negative)day-1 for (a), 5×10 value.4 J day- 1 -5 -1 -1 −5 4−1 -1 −1 4 −1 sula and Borneo (0 to 5◦ N and 100 to 105◦ E) extends to Contour5 for (b), interval2×10 s day is for 5 ×(c), 10and 5×10s J daydayfor (d). Thickfor black(a), contour 5 × 10 denotesJ day the value 6 of zero. The× gray− shade5 − is1 statistical−1 significance at 95% of confidence× 4 level for−1 the difference the northwestern coast of Borneo due to the enhanced vortic- for7 from(b) ,the 2 climatology.10 s day for (c), and 5 10 J day for (d). ity transport. At lower levels such as 925 hPa, this patch of Thick black contour denotes the value of zero. The grey shading is statistical significance at 95 % confidence level for the difference anomalous flux convergence further extends northeastward from the climatology. along the Borneo coast to around 115◦ E, 6◦ N. Averaged be- tween 0 and 4◦ N, the anomalous flux convergence is largest in the lower troposphere (∼ 700 hPa or larger) over the equa- ◦ energy in the lower troposphere and hence stabilizes the at- torial South China Sea between 100 and 110 E during SS mospheric column. (Fig. 3c), increasing the ambient absolute vorticity. Both the convergence of absolute vorticity and moist static Next, we examine the moist static energy, Em = g Z + energy fluxes over the equatorial South China Sea under such cp T + Lq (g: gravitational acceleration, Z: geopotential SS conditions may be favourable for spinning up the Bor- height, cp: specific heat at constant pressure, T : tempera- neo vortex and sustaining cumulus convection: the tendency ture, L: latent heat of condensation, q: specific humidity). to increase ambient absolute vorticity encourages stronger This quantity serves as an indicator for potential activity of vorticity generation by vortex stretching, while there is am- moist cumulus convection and associated rainfall in the Bor- ple moist static energy supply to sustain convection and neo vortices. The isobaric divergence of moist static energy latent heat release, which underlies the buoyancy of air ∇ · 0 flux anomalies at 1000 hPa during SS, defined as p (V Em) parcels causing vertical motion and vortex stretching. To test (where the prime denotes deviation from the climatologi- whether the SS conditions favour the formation of the Bor- cal value), has a similar pattern to that of absolute vortic- neo vortex, we conducted the semi-idealized experiment de- ity flux anomalies (Fig. 3a and b): the moist static energy scribed in Sect. 5. transported equatorward by the northeast monsoon converges over the equatorial South China Sea during SS. Note that 3.3 Rainfall and its diurnal cycle the anomalous flux convergence is most significant below 800 hPa (Fig. 3d). Thus, moist static energy is anomalously Intense rainfall over the western Maritime Continent in De- transported to the equatorial South China Sea by strong cold cember is prominent over the central South China Sea, the surges from the evaporative sources in northern South China eastern Malay Peninsula, northwestern coastal sea of Borneo Sea. This supply of moist static energy maintains moist con- and the Java Sea (Fig. 4a). The rainfall over these regions vection in the equatorial South China Sea where vertical mix- shows considerable variability with the cold surge activity ing by convective updrafts and downdrafts as well as net ver- (Fig. 4b and c). tical transport to the upper troposphere removes moist static www.atmos-chem-phys.net/14/4539/2014/ Atmos. Chem. Phys., 14, 4539–4562, 2014 4544 S. Koseki et al.: Borneo vortex and mesoscale convective rainfall

Fig. 4. (a) The climatology of daily accumulated rainfall obtained from TRMM 3B42 and anomalies during NS and SS from the climatology in (b) and (c), respectively. The black dotted lines denote statistical significance at 90 % confidence level. The boxes A–C in (a) are for the area averages in Fig. 5.

(mm/hour) (a) (b) (c) 1.3 1.3 1.3 Climatology 1.1 SS 1.1 1.1

NS 0.9 0.9 0.9

0.7 0.7 0.7

0.5 0.5 0.5

0.3 0.3 0.3

0.1 0.1 0.1 00 03 06 09 12 15 18 21 00 (LST) 00 03 06 09 12 15 18 21 00 00 03 06 09 12 15 18 21 00

Fig. 5. Three-hourly rainfall averaged over (a) the equatorial South China Sea (box A), (b) the South China Sea north of Borneo (box B), and (c) the Java Sea (box C) during NS (long-short dashed line), climatology (solid line), and SS (dashed line).

(a) ux (b) vy Over the land and coastal1 Fig seas,ure 5. the3hour rainfallly rainfall hasavera aged strong over (a) the eqEquatorial South China Sea (box A), (b)Eq the 2 diurnal cycle brought aboutSo byuth C sea/landhina Sea no breezerth of Bor circulationneo (box B), and (c) t2She Java Sea (box C) during NS (long-s2Short 3 dashed line), climatology (solid line), and SS (dashed line). (e.g. Mak and Walsh, 1976; Hadi et al., 2002; Joseph et al., 4S 4S 2008; Teo et al., 2011). In contrast, the diurnal cycle of the 6S 6S rainfall over the South China Sea has smaller amplitude es- 8S 8S pecially near the Equator (see the boxes in Fig. 4a and the 10S 10S solid lines in Fig. 5). This may be because the local SST di- 100E 104E 108E 112E 116E 120E 100E 104E 108E 112E 116E 120E urnal fluctuation is small over the open sea and the influence -1 -2.7 -2.1 -1.5 -0.9 -0.3 0.3 0.9 1.5 2.1 (*10e-5 s ) of the landmass in the form of land breeze is either weak or absent. During SS (NS) the rainfall over the equatorial South Fig. 6. The difference in horizontal convergence at 850 hPa (shad- China Sea is almost uniformly enhanced (reduced) through- ing) between SS and climatology for (a) zonal component and out the day (Fig. 5a). In contrast, the climatological rainfall (b) meridional component obtained from JRA25/JCDAS data. Figure 6. The difference in horizontal convergence at 850 hPa (shading) between SS and over the Java Sea has a large diurnal cycle and during SS climatology for (a) zonal component and (b) meridional component obtained from (NS) the rainfall is enhanced (reduced) the most in the morn- JRA25/JCDAS data. ing around 03:00–12:00 LST (06:00–09:00 LST), when the the westerly wind component, the northerly wind compo- rainfall is maximal (Fig. 5c). nent is clearly enhanced in the Java Sea (Fig. 6). Thus, it is Changes in the diurnal cycle under changing ambient wind dynamically consistent that we see enhanced rainfall in the have been reported in Qian et al. (2010) and Koseki et morning but not so much in the afternoon over the Java Sea al. (2013) over the Java Sea. The essential mechanism for (Fig. 5c). We surmise that during SS, the rainfall change over intense rainfall over the Java Sea during the boreal winter the Java Sea is the result of the interaction of the synoptic- monsoon demonstrated in Koseki et al. (2013) is that the scale circulation with the local-scale circulation and is un- seaward land breeze collides with the ambient wind which likely to be directly linked with the mechanisms underlying has a northerly component leading to the peak rainfall oc- the rainfall change over the South China Sea. In the same curring in the morning. During strong cold surges, unlike vein of thought, we shall not further investigate the changes in the rainfall over the eastern Malay Peninsula and near

Atmos. Chem. Phys., 14, 4539–4562, 2014 www.atmos-chem-phys.net/14/4539/2014/ S. Koseki et al.: Borneo vortex and mesoscale convective rainfall 4545 northwestern Borneo as the local-scale coastal circulation in (/s) (m/s) these areas are as strong as over the Java Sea. 6e-05 20

4e-05 16 12 4 Borneo vortex in JRA25 and JCDAS data 2e-05 8 0 Since rainfall is intensified the most over the South China 4 -2e-05 Sea during SS roughly where the absolute vorticity anomaly 0 exceeds 0.4 × 10−4 s−1 over (compare Figs. 2f and 4c), one 1981 1982 1983 1984 1985 1986 1987 1988 1989 may suspect that a physical relationship may exist between 6e-05 20 the absolute vorticity and rainfall activity. In particular, we 4e-05 16 are interested in those situations when the high absolute vor- 12 ticity manifests as a strong Borneo vortex. 2e-05 8 We define the Marine Borneo Vortex (MBV) Index as the 0 averaged absolute vorticity at 850 hPa averaged over an area 4 ◦ ◦ -2e-05 bounded by 0 to 4 N and 105 to 110 E covering the equa- 0 torial South China Sea between the Malay Peninsula and 1990 1991 1992 1993 1994 1995 1996 1997 1998 Borneo. This is because the absolute vorticity flux conver- 6e-05 20 16 gence is remarkable in this region (as indicated in Fig. 3a) 4e-05 and the cyclonic vortex tends to be induced there. The MBV 12 Index tends to be strong when the CS Index is strong (Fig. 7) 2e-05 8 and vice versa. The correlation coefficient between CS and 0 MBV indices is 0.35 in December and 0.49 in January, both 4 of which are statistically significant at the 95 % confidence -2e-05 0 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 level. This suggests that a strong cold surge which trans- ports high absolute vorticity from the north tends to spin up Fig. 7. The temporal sequences of daily CS Index (shade) and the a strong Borneo vortex over the equatorial South China Sea. Marine Borneo Vortex (MBV) Index (line) in December obtained from JRA25/JCDAS. Dark (light) shade is CS Index larger (smaller) This idea is supported by the dynamical experiments of Lim Figure 7. The temporal sequences of daily CS Index (shade) and the Marine Borneo vortex and Chang (1981). than(MBV climatology) Index (line) in D plusecemb oneer obt standardained from J deviationRA25/JCDAS of. Da CSrk (lig Index.ht) shade Theis CS Index dashedlarger (sm linesaller) denotethan clim theatolo climatologygy plus one stan plusdard d oneeviati standardon of CS In deviationdex. The da ofshed lines Next, we define a strong cold surge event and a strong ma- denote the climatology plus one standard deviation of MBV Index. The lines are disjoint between rine Borneo vortex event as a continuous period when the MBVyears. Index. The lines are disjoint between years. daily mean value of both the CS and MBV indices are larger than one standard deviation from their respective climatolog- the wind field is not axisymmetric in the composited Bor- ical mean. A strong cold surge is said to spin up a strong neo vortex, with strong northeasterlies in the northeast sector, Borneo vortex in the equatorial South China Sea only if the which is attributed to the strong cold surge. The maximum of two events overlap in time and the former starts at the same the absolute vorticity is located slightly northwest from the time or earlier than the latter. With such definitions, we found vortex centre and the ridge of high absolute vorticity extends that 58 out of the 133 days of SS identified previously are re- northeastward and, to a much lesser extent, westward and lated directly to the strengthening of the Borneo vortex in southeastward. The strongest convergence is at the north of the equatorial South China Sea. For three of these days, the the centre but the convergence is also relatively strong in the Borneo vortex is located north of 5◦ N and takes on a more southern, western and northeastern sectors (Fig. 8c) where elongated shape, which is rather exceptional. Therefore, sub- strong wind prevails (Fig. 8b). On the other hand, weak diver- sequent composite analyses use only the other 55 days of gence is located southeast of the composited vortex overlap- the diagnosed vorticity and divergence. Only 16 days of the ping with a patch of weakly negative absolute vorticity. The satellite rainfall data are used because of TRMM’s shorter pattern of intense rainfall is approximately consistent with temporal coverage. the pattern of strong convergence with a slight northeastward Figure 8a shows that strong Borneo vortices during SS displacement of the former from the latter. occur mostly around the west of Borneo near the Equator As the resolution of the JRA/JCDAS is low and vortices between 5◦ N and 2◦ S where tropical cyclones do not de- of different sizes and shapes are composited in Fig. 8b and c, velop (except for the singular case of the Typhoon Vamei) it is hard to discern any clear relation between absolute vor- The technical details of how the location of the vortex cen- ticity, wind convergence and rainfall in a Borneo vortex from tre is obtained are explained in Appendix A. This distribu- the above results. tion of vortex centres is similar to that obtained by Chang et al. (2003), although they identified Borneo vortices without regard to the strength of the cold surge. Figure 8b shows that www.atmos-chem-phys.net/14/4539/2014/ Atmos. Chem. Phys., 14, 4539–4562, 2014 4546 S. Koseki et al.: Borneo vortex and mesoscale convective rainfall

(a)

18N

12N

6N

1 3 5 2 1 2 3 2 2 1 1 2 2 2 1 1 Eq 4 2 5 6 3 1 1 1 1

6S

90E 100E 110E 120E 130E

(km) -5 -1 (km) (b) ( 10 s ) (c) (mm/day) 1200 1200 4 40

800 800 3 30

400 400 2 20

0 C 1 0 c 10

-400 0 -400 1

-1 -800 -800

-1200 -1200 -1200 -800 -400 0 400 800 1200 (km) -1200 -800 -400 0 400 800 1200 (km)

10 (m/s)

Fig. 8. (a) The location of Borneo vortex centre obtained from JRA25/JCDAS. The grid numbers give the number of days that the centre is located at each grid. (b) The composite1 F ofigu absolutere 8. (a) Th vorticitye location of (shaded)Borneo vortex and centr horizontale obtained from wind JRA25 (vector)/JCDAS. Th ate n 850umberhPa from JRA25/JCDAS around the 2 of each grid counts the number of day when the centre is located at each grid. (b) The composite centre of Borneo vortex. The white3 thick contour denotes the value of zero of absolute vorticity. (c) The composite of horizontal divergence − − of absolute vorticity (shaded) and horizontal wind (vector) at 850 hPa from JRA25/JCDAS (contour interval is 10 6 s 1) at 8504 hPaaroun fromd the cen JRA25/JCDAStre of Borneo vortex. T andhe whi hourlyte thick con rainfalltour denote (shaded)s the value of fromzero of a TRMMbsolute around the centre of the Borneo 5 vorticity. (c) The composite of horizontal divergence (contour, interval is 10-6 s-1) at 850 hPa vortex. Solid (dashed) line means positive6 from (negative)JRA25/JCDAS value.and hourl Cy ra denotesinfall (shade thed) fr centreom TRMM of ar theound compositethe centre of the Borneo Borneo vortex. 7 vortex. Solid (dashed) line means positive (negative) value. C denotes the centre of the 8 composite Borneo vortex. 5 Semi-idealized experiment for the Borneo vortex is able to generate a vortex, despite using initial and bound- ary conditions obtained from a set of SS events that includes Based on the CS and MBV indices obtained in the preced- days where no Borneo vortex forms is due to the anomalous ing section, Borneo vortices occur with a probability of 44 % convergence of the absolute vorticity flux over the equatorial (= 58 days out of 133 days) during strong cold surges. The South China Sea (Fig. 3a); by holding such lateral boundary presence of Borneo vortices may be a result of the enhanced conditions constant during the model simulation, we main- vorticity flux convergence in the lower troposphere discussed tain a strong cold surge in the model that keeps supplying in Sect. 3.2. In this section, we present the results of a numer- absolute vorticity and moist static energy over the equatorial ical experiment using NHM (see Sect. 2) designed to investi- South China Sea. Those 75 days where no Borneo vortices gate into this possibility as well as how rainfall is organized were generated in the real atmosphere are due to the lack of by the strong cold surge and the Borneo vortex. such persistent “lateral boundary forcings”. The integration period is 11 days and we set an expedi- 5.1 Experimental design ential initialization date–time of 15 December, 00:00 UTC for the model’s diurnally and seasonally dependent radia- We use the composite reanalysis fields for all 133 days of SS tion scheme. The model spins up effectively within the first as the model’s initial and (constant) boundary conditions to day of integration. Horizontal and vertical resolutions are represent the synoptic conditions under which a Borneo vor- 10 km × 10 km at the Equator using Mercator projection and tex forms over the equatorial South China Sea. Although one 40 layers in terrain-following coordinate. The centre of do- may use the composited fields from the 55 or 58 days where main is at (110◦ E, 2.5◦ N) and with 300 × 300 grid points Borneo vortices are actually present (see Sect. 4), we opt for (Fig. 9a). Realistic land–sea mask and terrain height are used. the earlier set of initial and boundary conditions since a Bor- Henceforth, we call this experiment the semi-idealized ex- neo vortex is spun up in our model using them. Furthermore, periment. We checked that there was no systematic trend this set of initial and boundary conditions better represents in total mass within the domain despite the prescription of the conditions during SS. The reason as to why the model boundary winds.

Atmos. Chem. Phys., 14, 4539–4562, 2014 www.atmos-chem-phys.net/14/4539/2014/ S. Koseki et al.: Borneo vortex and mesoscale convective rainfall 4547

(a) (b) 24 hour (c) 48 hour (d) 72 hour 15N Luzon 15N 15N 15N

9N 9N 9N 9N South China Sea Malay 3N Peninsula 3N 3N 3N

Sumatra Borneo

3S 3S 3S 3S Java Sea Java 9S 9S 9S 9S

99E 105E 111E 117E 123E 99E 105E 111E 117E 123E 99E 105E 111E 117E 123E 99E 105E 111E 117E 123E

(e) 96 hour (f) 120 hour (g) 144 hour (h) 168 hour 15N 15N 15N 15N

9N 9N 9N 9N

3N 3N 3N 3N

3S 3S 3S 3S

9S 9S 9S 9S

99E 105E 111E 117E 123E 99E 105E 111E 117E 123E 99E 105E 111E 117E 123E 99E 105E 111E 117E 123E

(i) 192 hour (j) 216 hour (k) 240 hour (l) 15N 15N 15N 4N ◎ 3N ◎ ◎ ◎ 2N 240h 9N 9N 9N ◎

1N ◎ ◎ 24h Eq ◎ ◎ 3N 3N 3N ◎ 1S 2S 3S 3S 3S 3S 103E 105E 107E 109E 111E

9S 9S 9S

99E 105E 111E 117E 123E 99E 105E 111E 117E 123E 99E 105E 111E 117E 123E

20 (m/s)

Fig. 9. (a) The domain of the semi-idealized experiment. (b)–(k) 24-hourly temporal evolution in the simulated Borneo vortex shown by horizontal wind at 850 hPa from 24 to 240 h after initialization. (l) shows the 24-hourly cyclone track.

1 Figure 9. (a) The domain of the semi-idealized experiment. (b)-(k) 24 hourly temporal evolution 5.2 Simulated Borneo vortex At 120 h, the vortex strength is enhanced more significantly. 2 in the simulated Borneo vortex shown by horizontal wind at 850 hPa from 24 to 240 hours from From 144 to 192 h, the Borneo vortex moves slowly north- 3 initialization. (l) shows the 24-hourly cyclone track. westward. This well-developed vortex has a spatial scale be- Figure 9b–k show the formation and westward movement of tween a few hundred to a thousand kilometres, i.e. at the the vortex over the equatorial South China Sea in the semi- meso-α scale. This meso-α cyclone is unlike a tropical cy- idealized experiment. At 24 h, meridional wind at 850 hPa is clone which is more intense and has distinct dynamics. Be- almost southward over the equatorial South China Sea and tween 216 and 240 h, the cyclone reaches over the Malay the Borneo vortex has barely formed yet. Between 48 and Peninsula and weakens. Figure 9l illustrates the 24 hourly 72 h, weak southerly is detected around 108 to 109◦ E and ◦ ◦ track of the cyclone centre (defined in Appendix A) from 1.5 S to 0 N and cyclonic circulation is generated around 24 to 240 h. the Equator over the sea. This cyclonic circulation is intensi- Figure 10a and b illustrate time sequences of area- fied by 96 h. At this stage, the cyclonic circulation resembles averaged absolute vorticity at 850 hPa and hourly rainfall the typical Borneo vortex and the vortex is quasi-stationary. www.atmos-chem-phys.net/14/4539/2014/ Atmos. Chem. Phys., 14, 4539–4562, 2014 4548 S. Koseki et al.: Borneo vortex and mesoscale convective rainfall

within 100 km of the cyclone centre. The absolute vorticity ( s-1 ) (a) −4 −1 0.0006 is less than 10 s until 50 h and becomes gradually inten- Mature 0.0005 sified until around 140 h. After that, the cyclone is more or Decaying less mature. From 190 h, it weakens rapidly and remains as a 0.0004 Developing 0.0003 weak vortex after 223 h. Rainfall shows a clear diurnal cycle 0.0002

(roughly consistent with the diurnal cycle in Fig. 4a) and the 0.0001 diurnal peak rainfall intensity appears to follow the trend of 0 the absolute vorticity of the meso-α cyclone. Based on these 40 60 80 100 120 140 160 180 200 220 240 260 (hour) temporal sequences of the cyclone, we can divide the life cy- cle of the cyclone into three stages: a developing stage (50 to (mm/h) (b) 140 h), a mature stage (141 to 190 h), and a decaying stage 20 (191 to 223 h). 16 Figure 11 shows the horizontal distribution of temporal 12 mean of hourly accumulated rainfall around the cyclone cen- 8 tre at each stage. In the developing stage (Fig. 11a), rainfall is 4 intense just west of the cyclone centre and extends northward 0 ∼ 150 km. In the mature stage (Fig. 11b), the rainfall adja- 40 60 80 100 120 140 160 180 200 220 240 260 cent to the cyclone centre becomes more intense and a well- Fig. 10. The time sequences of (a) absolute vorticity at 850 hPa and developed comma-shaped rainband extends from the centre (b) hourly rainfall averaged within a radius of 200 km of the centre northward, northeastward and then eastward to reach a dis- 1 Figure 10. The time sequences of (a) absolute vorticity at 850 hPa and (b) hourly rainfall of simulated2 averaged meso-within a radiusα cyclone. of 200 km of Thethe centre grey of simulated bars meso- denoteα cyclone. the The grey time bars from tance of 250 km from the centre. This comma-shaped rain- 00:003 todenote 12:00 the time LST. from 00 to 12 LST. band is located only in the northern part of the cyclone and its localization is discussed in Sect. 6. In the decaying stage (Fig. 11c), the rainfall around the centre is reduced signifi- The fact is that the Borneo vortices for the 16 days com- cantly and the distribution is relatively widespread but still posited in the TRMM data record show different horizon- mostly in the northern part of the cyclone. Throughout the tal size, shape and centre locations and the rainfall also has three stages of the simulation, no “eye” formation occurs, much spatial variance around the vortex centre. Conversely, unlike in typhoons which form in higher tropical latitudes our simulated vortex is one idealized case and has quite a and in the Typhoon Vamei which formed in this region. clear feature of a rainband. Nonetheless, some qualitative Compared with the composite of the Borneo vortex in the features, such as the relatively strong rainfall in the strong TRMM data set (Fig. 8c), the rainfall is much more localized convergence zone northeast of the vortex centre, are seen in in the northern and northeastern sector, and is one order of both our simulation and the observations. magnitude more intense in our semi-idealized experiment. This is expected from the higher resolution of the model 5.3 Vertical structure of the meso-α cyclone which is able to capture the meso-β-scale dynamics. On the other hand, TRMM rainfall estimates could be too low due to We explore the structure of the axisymmetric vertical struc- interfering thick and widespread clouds in the Borneo vortex ture of the meso-α cyclone by averaging the atmospheric during data retrieval. Moreover, no two actual Borneo vor- variables at equal distances from the cyclonic centre at the tices have exactly matching intense rainfall regions and so mature stage, when the cyclone is most intense. Figures 12a a composite always reflects lower rainfall magnitudes than and b illustrate the axisymmetric tangential and radial winds. single vortices. Tangential velocities beyond 24 m s−1 are seen just within a The southern patch of relatively intense rain in the TRMM 50 km radius, which is comparable to that of a typical ty- composite (Fig. 8c) is not seen in the model simulation phoon over the tropical Pacific Ocean (Frank, 1977). How- (Fig. 11b). This may be because the centres of actual Bor- ever, such strong winds are limited to the lower troposphere neo vortices are distributed rather widely around the Equator (below ∼ 750 hPa) and the tangential wind rapidly weak- in Fig. 8c and so the southern part of some vortices (i.e. 400– ens in the upper troposphere (< 20 m s−1 above ∼ 600 hPa). 800 km south of the centre) spans the Java Sea where rain- Thus, the meso-α cyclone in our experiment is shallower fall arises mainly due to local land breeze circulation (see compared to a typical typhoon where winds remain quite Sect. 3.3) and is captured in the composite. But in the model, strong even in the upper troposphere (Frank, 1977). There the vortex centre lies between 2 and 3◦ N in the mature stage is radial inflow mainly below ∼ 900 hPa and radial outflow (from 144 to 192 h in Fig. 9l) and the simulated vortex is mainly above radial ∼ 300 hPa. The distribution and mag- about 2–3 times smaller. So the southern part of the simu- nitude of radial wind are similar to those of typical tropi- lated vortex (i.e. 200–300 km south of the centre) does not cal typhoons over the Pacific Ocean (Holland and Merrill, reach the Java Sea and therefore there is no southern intense 1984). The wind speeds are much stronger and the spatial rain patch. scale is much smaller in our simulation than in the composite

Atmos. Chem. Phys., 14, 4539–4562, 2014 www.atmos-chem-phys.net/14/4539/2014/ S. Koseki et al.: Borneo vortex and mesoscale convective rainfall 4549

(km) (a) (b) (c) 500 500 500

300 300 300

100 100 100 C C C -100 -100 -100

-300 -300 -300

-500 -500 -500 -500 -300 -100 100 300 500 (km) -500 -300 -100 100 300 500 -500 -300 -100 100 300 500

1 5 9 13 17 21 25 29 33 37 41 45 (mm/hour)

Fig. 11. The temporal mean of hourly rainfall around the centre of simulated meso-α cyclone in the (a) developing stage, (b) mature stage, and (c) decaying stage. C denotes the centre of composite meso-α cyclone.

(hPa) (a) (b) 100 100 6 3 300 300 2 1 Figure 11. The tempor9al mean of hourly rainfall around the centre of simulated meso-α cyclone 2 in the (a) developing stage, (b) mature stage, and (c) decaying stage. C denotes the centres of 3 co500mposite meso-α cyclone. 500

700 700 15 21 -2 900 900 -6 50 150 250 350 450 (km) 50 150 250 350 450

-5 (K) (10 kg/kg) (c) (d) 100 100 2.1 9 1.5 300 -0.02 300 7 0.9

500 5 500 0.3

279 -0.3 700 -0.01 3 700 285 -0.9

1 291 900 900 -1.5 297 50 150 250 350 450 50 150 250 350 450 (hPa) (e) 1010

1006

1002

998

50 150 250 350 450

Fig. 12. The pressure and radial sections1 Figure of 12 axisymmetric. The pressure and radial components sections of axisymmetric of (a) tangential components wind,of (a) tangential(b) radial wind, wind, (c) mixing ratio of cloud water 2 (b) radial wind, (c) mixing ratio of cloud water (liquid plus−1 ice, shade) and vertical velocity (liquid plus ice, shade) and vertical3 velocity(contour, interval (contour, is 0.005 interval hPa/hour), is(d) 0.005temperature hPa (contour) h ), and(d) temperaturetemperature deviation (contour) from and temperature deviation from radial mean (shaded). The white thick4 radial contours mean (shaded). in (d) Theare white the thickvalue contours of zero in (d) of are temperature the value of zero deviation. of temperature(e) The axisymmetric component of sea 5 level pressure. deviation. (e) The axisymmetric component of sea level pressure.

Borneo vortex from the JRA25/JCDAS (not shown), proba- stage. One maximum of cloud water is detected between bly because the global model on which the JRA25/JCDAS is 900 and 500 hPa around the radius of 50 km. This maximum based lacks the resolution to support the downscale cascade of cloud water is associated with the deep convection and of kinetic energy in the mesoscale which serves to strengthen heavy rainfall adjacent to the cyclone centre (Fig. 11b). Our and tighten the vortex. simulated cyclone does not have a cloud-free region in the Figure 12c shows the axisymmetric component of mixing centre and the surface pressure depression is not strong, in- ratio of cloud liquid water plus cloud ice (shaded) and verti- dicating that the structure of our meso-α cyclone (the Bor- cal velocity (contour) around the cyclone centre in the mature neo vortex) is different from that of typical mature typhoons, www.atmos-chem-phys.net/14/4539/2014/ Atmos. Chem. Phys., 14, 4539–4562, 2014 4550 S. Koseki et al.: Borneo vortex and mesoscale convective rainfall although the intensity of radial and tangential wind in the and (4) tilting of horizontal vortex tubes (TILT). The cyclone lower troposphere is comparable to them (e.g. Frank, 1977; velocity only affects horizontal advection and hence the time Holland and Merrill, 1984). The location of the maximum of tendency in the co-moving frame. vertical velocity is roughly consistent with that of cloud wa- Figures 13a–e show that HADV and VADV contribute ter, especially after making allowance for possible outward negatively to the cyclone development. On the other hand, displacement of hydrometeors due to centrifugal forces in the terms of STRC and TILT are positive. In particular, STRC the cyclonic circulation. This strong upward motion is con- is more dominant and is mainly responsible for spinning sistent with the warm anomaly (Fig. 12d) due to latent heat up the cyclone. The net forcing shows maximum positive release which can be attributed to the deep cumulus con- tendency west of the absolute vorticity maximum and the vection around the vortex centre. While the axisymmetric cyclone centre, as horizontal convergence maximizes there vertical velocity shows negative values (upward), weak non- leading to strong vertical vortex stretching. As a result, the axisymmetric positive vertical velocity (downward) can be cyclone migrates westward with time. detected around the centre as part of the local organization While the sign of each term is roughly the same in the of convective downdrafts (not shown). The other maximum mature stage (Fig. 13f–j) as in the developing stage, the of cloud water (mostly in the ice phase) is seen between magnitude of every term, especially STRC, is much larger. 300 and 100 hPa and has a large radial extent from 100 to Additionally, a positive tendency is seen in the HADV and 350 km. This is because the radial wind at this level is cen- VADV denoting that advection is bringing high-vorticity air trifugal and the cloud ice is advected outward. Thus, this up- to the cyclone centre from the northwest and from lower lev- per tropospheric cloud can be categorized as one of the anvils els where the vorticity generation is stronger. TILT shows associated with cumulus convection. Moreover, the horizon- a north–south dipole structure resulting from the maximum tal distribution of the cloud liquid water and ice is actually rising motion located at the centre of the dipole under am- asymmetric around the meso-α cyclone (not shown). bient easterly wind shear. The net positive forcing still lies Meanwhile, the depression in sea level pressure (SLP) at west of the absolute vorticity maximum and the cyclone cen- the cyclone centre corresponds well to the positive temper- tre as that is where vertical vortex stretching is the strongest, ature anomaly aloft as is expected from hydrostatic balance but a negative tendency is also found in the southwestern and (Fig. 12e). The mesoscale simulation indicates that the cen- southern sector of the cyclone due to the tilting effect. tral depression could be as much as 13 hPa lower than the am- Higher up in the atmosphere (600 hPa) in the mature stage bient SLP, which is an order of magnitude greater than what (Fig. 13k–o), HADV, VADV and STRC are largely opposite is recorded in the JRA25/JCDAS data set at synoptic scale to those at 850 hPa: absolute vorticity is advected upward and (not shown). This is reasonable because most of the pressure westward from the lower levels while horizontal divergence drop occurs within 150 km from the cyclone centre, which is is destroying the vorticity. TILT still has a dipole structure, below the limit of the resolution of the JRA25/JCDAS data but in the east–west rather than north–south direction. The set. net forcing is still positive west of the vorticity maximum and the cyclone centre mainly due to vertical advection from the lower levels. At even higher levels (∼ 250 hPa, not shown), 6 Analysis and discussion of the numerical experiment while the tendency of each forcing term is similar to those at 600 hPa, the net forcing is only weakly positive. 6.1 Growth and maintenance of the meso-α cyclone In summary, vortex stretching mainly occurs at low levels and the ambient cross-isobaric absolute vorticity is rapidly To investigate what mechanism leads to the growth of the intensified and advected to the mid-troposphere where it is meso-α-scale cyclone, we examine the absolute vorticity dy- destroyed by isobaric divergence. This is the primary reason namics in this section. for the spin-up and sustenance of a meso-α cyclone with the Appendix B shows that in the reference frame co-moving shallow vertical structure. The tilting of vortex tubes gener- with the cyclone centre, the tendency equation for cross- ates both positive and negative cross-isobaric absolute vor- isobaric absolute vorticity ζ is a ticity and this serves only to re-organize the vorticity distri- ∂ζ ∂ζ ∂ω ∂v ∂ω ∂u bution on the isobaric surface in the mature stage. a =−(V −V )·∇ ζ −ω a −ζ D− − , (1) ∂t H C p a ∂p a ∂x ∂p ∂y ∂p 6.2 Dynamics–thermodynamics relationship in the where V H and ∇p are as defined in Sect. 3.2, V C ≡ (uC, meso-α cyclone core vC) is the cyclone velocity estimated from the cyclone track in Fig. 9l, ω is the vertical velocity in pressure coordinate, The previous section showed that our simulated vortex is and D is the isobaric divergence. The right-hand side of maintained by the stretching term adjacent to the cyclone Eq. (1) comprises (from left to right) the following forcing core (Fig. 13h). The intense stretching term is consistent terms: (1) horizontal advection (HADV), (2) vertical advec- roughly with the strong rainfall close to the centre (Fig. 11b), tion (VADV), (3) stretching of vertical vortex tubes (STRC), indicating the relationship between the vortex tube stretching

Atmos. Chem. Phys., 14, 4539–4562, 2014 www.atmos-chem-phys.net/14/4539/2014/ S. Koseki et al.: Borneo vortex and mesoscale convective rainfall 4551

HADV VADV STRC TILT Net (km) (a) (b) (c) (d) (e) 100 100 100 100 100 ( s-1 )

0.001 60 60 60 60 60 0.0008 20 20 20 20 20 0.0006 C C C C C

-20 -20 -20 -20 -20 0.0004 Developing 850 hPa 0.0002 -40 -40 -40 -40 -40

-100 -100 -100 -100 -100 -100 -60 -20 20 60 100 -100 -60 -20 20 60 100 -100 -60 -20 20 60 100 -100 -60 -20 20 60 100 -100 -60 -20 20 60 100

(h) (i) (j) -1 100 (f) 100 (g) 100 100 100 ( s )

0.0022 60 60 60 60 60 0.0018 20 20 20 20 20 0.0014 C C C C C -20 -20 -20 -20 -20 0.001 Mature 850 hPa

0.0006 -40 -40 -40 -40 -40 0.0002 -100 -100 -100 -100 -100 -100 -60 -20 20 60 100 -100 -60 -20 20 60 100 -100 -60 -20 20 60 100 -100 -60 -20 20 60 100 -100 -60 -20 20 60 100

(k) (o) 100 100 (l) 100 (m) 100 (n) 100 ( s-1 ) 60 60 60 60 60 0.0016

20 20 20 20 20 0.0012 C C C C C 0.0008 -20 -20 -20 -20 -20

Mature 600 hPa 0.0004

-40 -40 -40 -40 -40

-100 -100 -100 -100 -100 -100 -60 -20 20 60 100 -100 -60 -20 20 60 100 -100 -60 -20 20 60 100 -100 -60 -20 20 60 100 -100 -60 -20 20 60 100

Fig. 13. The temporal mean of (a) HADV, (b) VADV, (c) STRC, (d) TILT and (e) net forcing in Eq. (1) in the developing stage at 850 hPa. The solid (dashed) lines denote positive (negative) values. The contour interval is 10−6 s−2. (f)–(j) and (k)–(o) are the same as for (a)–(e), but for the mature stage at 850 and 600 hPa, respectively. The shading shows absolute vorticity. C denotes the centre of meso-α cyclone.

(dynamics) and1 theFig moisture 13. cumulus The temp convectionoral mean of (thermody- (a) HADV, (b) V6.3ADV, What (c) ST maintainsRC, (d) TIL theT an strongd (e) ne convergencet forcing in the namics). Figure2 14ain t showshe equa thetion vertical of (1) in profile the dev ofelo thepin mixingg stage at 850 hPa. meso-The soαlidcyclone? (dashed) lines denote positive -6 -2 ratio of five hydrometeors3 (negative) usedvalue ins. T NHMhe con averagedtour interv overal is the10 s . (f)-(j) and (k)-(o) are same as for (a)-(e), but 4 black box in Fig. 13hfor th ine thematu maturere stage stage. at 850 Basically, hPa and 6 the00 h rainPa, respeWhilectively.the The low-level shading sh vortexows ab stretchingsolute vortic dueity. to isobaric con- 5 C denotes the centre of meso-α cyclone. The black box in (h) denotes the area for Fig. 14. and snow are generated mostly in the lower and upper tro- vergence is mainly responsible for the intensification of the posphere, coinciding with the intense rainfall at the surface. cyclone, what process is responsible for the maintenance of This implies that the latent heat released in the formation of the convergence around the cyclone centre? the rain and snow helps in the formation of the warm core by Appendix B shows that the divergence tendency equation overcoming the effect of cooling due to adiabatic expansion in the frame co-moving with the cyclone centre can be writ- as air parcels rise in the cyclone centre (Fig. 12d). ten as The vertical advection of the potential temperature, shown ∂D  ∂  in Fig. 14b, cools the troposphere down. This cooling is due =−2(V H−V C)·∇pD+∇p· − [ω (V H−V C)] to the adiabatic expansion associated with the ascending mo- ∂t ∂p tion close to the centre shown in Fig. 12c and it is present 3 1 − D2−2λ2+ ζ 2−∇28−β (u−u ), (2) in order to compensate the diabatic heating due to the con- 2 2 a p c densation/deposition/freezing processes. The upward motion around the centre, in turn, stretches the vortex tube of the where λ is the deviatoric strain (deformation less the effect of meso-α cyclone, which contributes significantly to the cy- the horizontal divergence) rate of the isobaric flow field and β clone maintenance (Fig. 13). is the gradient of the Coriolis parameter. The other symbols are as defined in Sect. 6.1. The right-hand side of Eq. (2) www.atmos-chem-phys.net/14/4539/2014/ Atmos. Chem. Phys., 14, 4539–4562, 2014 4552 S. Koseki et al.: Borneo vortex and mesoscale convective rainfall

(hPa) (hPa) The net convergence tendency (Fig. 15h) is due mainly to (a) (b) 100 100 SELF (Fig. 15g). However, the self-enhancement of conver- 200 200 gence relies on the existence of net convergence in the first

300 300 place. In other words, if the initial divergent field is zero, the SELF term is also zero. We suggest that deviatoric strain 400 400 provides the original seed for the convergence tendency in 500 500 the cyclone because STRN is the only negative-definite forc- 600 600 ing in Eq. (2) and it is strongest where SELF and the net 700 700 convergence tendency are the strongest (see Fig. 15f–h). We

800 800 also note that the location of the maximum of the SELF term shows approximate agreement with the location of the hy- 900 900 drometeors in Fig. 14 (i.e. the box shown in Fig. 13h), in- 1000 1000 0 0.0004 0.0008 0.0012 0.0016 (kg/kg) -0.006 -0.004 -0.002 0 (K/s) dicating that the upward motion forced by the diabatic heat- ing from the condensation/deposition/freezing of hydrome- Fig. 14. The vertical profiles of (a) the mixing ratio of five hy- teors contributes to the strong low-level convergence for the drometeors (black dashed: cloud water, red: rain, green: cloud ice, SELF term to act as well. Therefore, the process of seeding blue: graupel, cyan: snow, and black solid: sum of the five hydrom- eteors) and (b) the vertical advection of the potential temperature, by STRN followed by super-exponential growth by positive feedback through SELF and enhanced by the latent heat re- −Figureω(∂θ/∂p) 14. The vertical, averaged profiles of ( overa) the mixing the blackratio of five box hydrometeors in Fig. (black 13h. dashed: cloud water, red: rain, green: cloud ice, blue: graupel, cyan: snow, and black solid: sum of the five lease from moist convection near the cyclone centre is how hydrometeors) and (b) the vertical advection of the potential temperature averaged over the black box in Fig. 13h. the meso-α cyclone is generated and maintained. The net convergence and net divergence tendencies appear comprises seven forcing terms: twice the horizontal advec- to be evenly matched in Fig. 15h, consistent with the weak tion of divergence (HADD), divergence of the horizontal net growth of the cyclone in the mature stage. In the devel- stress due to vertical transport of momentum (VMOM), self- oping stage, although each term is weaker than in the ma- enhancement of convergence or self-inhibition of divergence ture stage, their distributions are approximately the same and (SELF), destruction by deviatoric strain (STRN), generation the net forcing of convergence is stronger than that of di- by centrifugal acceleration due to local rotation (CENT), di- vergence, resulting in the overall growth of convergence in vergence of the pressure-gradient force denoted by the Lapla- the cyclone (not shown). Therefore, we suggest that the Bor- cian of pressure field (LAP), and beta-effect on zonal wind. neo vortex is spun up primarily by the low-level convergence The beta-effect was found to be much weaker than the other (Sect. 6.1) which is originally forced by the deviatoric strain. terms and will not be shown here. The subsequent growth and maintenance of vortex is mainly Figure 15 shows the horizontal distribution of the first due to the self-enhancement effect of convergence. The devi- six forcing terms in Eq. (2) at 850 hPa in the mature stage. atoric strain is possibly induced by the confluence of synop- Here, positive (negative) values in the figure indicate the di- tic cold surge and the cyclonic flow near the cyclone centre, vergence (convergence) tendency, which can be interpreted which is the subject of the next section. loosely as consistent with downward (upward) motion in the free atmosphere above if there is negligible vertical motion 6.4 How is the comma-shaped rainband organized in at the surface. The beta-effect is much weaker than the other the meso-α cyclone? terms and is not shown. CENT and LAP are very strong but have similar distributions of opposite signs (Fig. 15a The northern and northeastern sweep of the comma-shaped and b). They represent a basic balance between the out- rainband away from the cyclone centre has already been ward centrifugal and inward pressure gradient forces in a noted (Fig. 11). Here, we examine the reason for such de- cyclone, although some residual local divergent tendency is parture from axisymmetry. seen (Fig. 15c) which explains the net divergent tendency The most intense convergence is located near the cyclone (Fig. 15h). An approximate balance is also seen between centre to the northwest (Fig. 16a) with the intense comma- HADD and VMOM (Fig. 15d and e). This means that the head rainfall and strong convergence also seen in the northern convergence and divergence generated by horizontal stress and northeastern sectors, consistent with the location of tail northwest of the vortex centre (Fig. 15e) are mostly re- part of the comma-shaped rainband (Fig. 11b). Conversely, moved by (twice) the advection from the northeasterly wind a divergence zone is seen in the southern and eastern sectors (Fig. 15d). The dipole structure of VMOM has its origin adjacent to the cyclone centre. The zonal (∂u/∂x) and merid- as follows: at low levels, the upward flow maximizes near ional (∂v/∂y) contributions to divergence (Fig. 16b and c) ∼ (x, y) = (−30 km, 10 km) where the wind is northeasterly; both show quadrupole structure within 150 km from the ∼ this creates a local maximum in southwesterly stress (since cyclone centre. Beyond 150 km, the zonal and meridional the horizontal momentum is lost to mid-levels) and leads to convergence cells are stronger in the northwestern and north- stress convergence upwind and stress divergence downwind. eastern sectors, respectively.

Atmos. Chem. Phys., 14, 4539–4562, 2014 www.atmos-chem-phys.net/14/4539/2014/ S. Koseki et al.: Borneo vortex and mesoscale convective rainfall 4553

(km) (a) (b) (c) (d) 100 100 100 100 CENT LAP CENT+LAP HADD

60 60 60 60

20 20 20 20 C C C C -20 -20 -20 -20

-60 -60 -60 -60

-100 -100 -100 -100 -100 -60 -20 20 60 100 -100 -60 -20 20 60 100 -100 -60 -20 20 60 100 -100 -60 -20 20 60 100 (km) (f) (g) (h) 100 (e) 100 100 100 VMOM STRN SELF (c)+(d)+(e)+(f)+(g)

60 60 60 60

20 20 20 20 C C C C -20 -20 -20 -20

-60 -60 -60 -60

-100 -100 -100 -100 -100 -60 -20 20 60 100 -100 -60 -20 20 60 100 -100 -60 -20 20 60 100 -100 -60 -20 20 60 100

-6 -4 -2 0 ( 10e-4 s-1 )

Fig. 15. The temporal mean of (a) CENT, (b) LAP, (c) CENT+LAP, (d) HADD, (e) VMOM, (f) STRN, (g) SELF, and (h) net forcing of divergence in Eq. (2) in the mature stage at 850 hPa. Solid (dashed) lines denote positive (negative) values. The contour interval is 10−6 s−2. The shading shows divergence. C denotes the centre of meso-α cyclone.

In a typical snapshot of the meso-α cyclone in the mature boundary layer which is one mechanism proposed to be re- stage (Fig. 16d), finer structures in the form of three or four sponsible for the generation of eyewalls in tropical cyclones rainfall cells 10–1001 kmF inigu sizere 15 can. The be te seenmpora alongl mean theof ( commaa) CENT, (b) LandAP, ( typhoonsc) CENT+L (e.g.AP, (d) Ooyama, HADD, (e) 1964; VMOM Yamazaki,, 1983). The 2 (f) STRN, (g) SELF, and (h) net forcing of divergence in the equation (2) in the mature stage at tail of the rainband.3 Intense850 hP cellsa. So oflid rainfall(dashed) canlines beden alsoote po seensitive (negCoriolisative) valu accelerationes. The contou isr in tooterv weakal is 10 near-6 s-2 the. Equator to support at the comma head4 at otherThe sh timesading s inho thews d northwesternivergence. C den sectorotes the centrthee of Ekmanmeso-α c pumpingyclone. mechanism proposed therein. Moreover, near the cyclone centre (not shown). if surface friction were important, there would be greater In the comma head, a marked change of the wind direction symmetry around the vortex as friction acts in the same way can be identified: with reference to the red box in Fig. 16d, in all directions, but Fig. 16a and d show that the southern while the zonal component of wind west of −100 km is sector of the vortex is clearly divergent instead of convergent. small, it is easterly and quite intense from −100 to 0 km In the northeastern sector in Fig. 16d, confluence of op- as part of the matured cyclonic flow. This change in direc- positely directed meridional winds is mainly responsible for tion results in the strong convergence (Fig. 16a–c) and in- the meso-β-scale convergence and rainfall evidenced by the tense comma-head rainfall in the northwestern sector around co-location of the zero meridional wind line and the comma the centre (Fig. 15). This strong convergence at the comma tail of the rainband. In fact, the intensities of the convergence head is also responsible for maintenance of the meso-α cy- and rainfall in the comma tail are comparable with those in clone itself through vortex stretching as shown in Fig. 13. the comma head. This fact is not apparent in the temporal While the intense easterly within −100 to 0 km is attributed mean picture (Figs. 11b and 16a) because the zero merid- mostly to the cyclonic flow itself, the main part of wind in ional wind line is not stationary relative to the cyclone centre the outer region (< −100 km) seems to come from the syn- (Fig. 16e). In other words, the meso-α-scale organization of optic background (larger-scale) wind rather than the meso-α the northeasterly surge and southeasterly cyclonic wind in cyclonic flow. This is because the northeasterly surge wraps the northeastern sector is the root cause of the rainband’s lo- itself around the cyclone core. The confluence zone of the cy- cation and the departure from axisymmetry. clonic flow and the synoptic cold surge generates and main- Figure 17 analyses the meridionally averaged dynamic tains the convergence by the seeding–positive feedback pro- and thermodynamic profiles in the northwestern sector ad- cess due to the STRN and SELF terms in Fig. 15, aided by jacent to the cyclone centre in the mature stage. Below the latent heat release by the intense cumulus convection that 800 hPa, the sharp change of zonal wind strength is evident produces the rainfall at the comma head. between −80 to −20 km and strong upward flow is gener- The above mechanism of generating strong convergence ated above that region (Fig. 17a). Following the upward mo- is different from frictional convergence in the atmospheric tion, positive (indicating relatively warm and wet) anomaly

www.atmos-chem-phys.net/14/4539/2014/ Atmos. Chem. Phys., 14, 4539–4562, 2014 4554 S. Koseki et al.: Borneo vortex and mesoscale convective rainfall

(km) (a) 500 500 (b) 500 (c)

400 400 400

300 300 300

200 200 200

100 100 100

0 C 0 C 0 C

-100 -100 -100

-200 -200 -200

-300 -300 -300

-400 -400 -400

-500 -500 -500 -500 -300 -100 100 300 500 (km) -500 -300 -100 100 300 500 -500 -300 -100 100 300 500

-0.0006 -0.0004 0.0002 0 0.0001

20(m/s)

(d) (e) 500 500

400 400

300 300

200 200

100 100

0 C 0 C

-100 -100

-200 -200

-300 -300

-400 -400

-500 -500 -500 -300 -100 100 300 500 -500 -300 -100 100 300 500

1 10 20 30 40 50 60 70 80 (mm/hour) 40 (m/s)

Fig. 16. (a) The temporal mean of horizontal divergence (shade) and horizontal wind (vector) in the mature stage at 850 hPa. This horizontal convergence is decomposed1 Figu intore 1(b)6. (zonala) The andtemp(c)orameridionall mean of ho components.rizontal diverg(d)encThee (sh snapshotade) and h ofor horizontalizontal wind divergence (vector) (white contours), hourly rainfall (shade), horizontal2 in t windhe ma (vector),ture stage andat 85 the0 hP zeroa. Th meridionalis horizontal wind conve linergen (blackce is dec contour)omposed atint 850o (b) hPa zon atal a 140nd ( hc) after the initialization. The white contours have3intervals meridi ofona 0.0002l compo s−ne1n.ts Solid. (d) T (dashed)he snaps lineshot of denote horizon positivetal diver (negative)gence (whit values.e contou(e)rs),The hour zeroly meridional lines plotted 4 rainfall (shade), and horizontal wind (vector), and the zero meridional wind line (black contour) at hourly intervals in the mature stage. C denotes the centre of meso-α cyclone. The red and black boxes in (d)-1 are the area for pressure– 5 at 850 hPa at 140 hours after the initialization. The white contours have intervals of 0.0002 s . longitude section in Fig.6 17Soli andd (d zoomedashed) lin regiones deno inte Fig.posit 19,ive respectively.(negative) values. (e) The zero meridional lines plotted at 7 hourly intervals in the mature stage. C denotes the centre of meso-α cyclone. The red and black 8 boxes in (d) are the area for the pressure-longitude sections in Fig. 17 and the zoomed region in 9 Fig. 19, respectively. in equivalent potential temperature, θe, reaches the upper tro- convective available potential energy (CAPE) is the total posphere (Fig. 17b), which creates the warm core as already work done on a near-surface air parcel as it ascends to the shown in Fig. 12c and d. Conversely, the negative (indicating LNB, a higher LNB means more work is done on the air relatively cool and dry) anomaly in θe is found in the outer parcel. Thus, CAPE follows the same pattern as LNB and it region (< −100 km), below 800 hPa. The negative anomaly determines the pattern of rainfall parameterized by the Kain– is consistent with the northerly flow and hence weak zonal Fritsch cumulus scheme as well as that captured by the mi- wind, indicating that the air mass in the outer region likely crophysics scheme due to explicit convection simulated in originates from the synoptic cold surge from the higher lati- the model (Fig. 17c). The rainfall east to the cyclone centre tudes that has not had time to warm up or moisten yet. is smaller compared to the west in spite of relatively high With reference to Fig. 17c, 60–70 % of the total rainfall LNB and CAPE. This may be because divergence is found around the centre is contributed by the cloud microphysics east of the centre (Fig. 15h) which indicates descending air scheme in the model. So, to a large extent, the strong up- which suppresses cumulus convection. ward motion induces condensation and rain formation which The location of the lowest lifting condensation level (LCL) in turn releases latent heat which has a positive feedback on of about 150 m around −10 km implies additionally that the the rising motion. Consequently, the warm core of the cy- near-surface air forced upwards by low-level frontal conver- clone is maintained, as shown in Fig. 12d as well. gence does not need to rise very high to achieve positive With reference to Fig. 17d, level of neutral buoyancy buoyancy. In contrast, between −200 and −140 km, the LCL (LNB) has a maximum to the west of the centre. Since reaches around 300 to 400 m. This makes it easier to realize

Atmos. Chem. Phys., 14, 4539–4562, 2014 www.atmos-chem-phys.net/14/4539/2014/ S. Koseki et al.: Borneo vortex and mesoscale convective rainfall 4555

(hPa) (a) (hPa) (b)

-12 200 200 1

-15 2 400 400 -18

-21 600 600

800 800 -3

3 6 9 1000 1000 -200 -140 -80 -20 40 100 (km) -200 -140 -80 -20 40 100 (km)

-0.10 -0.08 -0.06 -0.04 -0.02 0 (hPa/hour) (m) (m) (mm/hour) (c) (J/kg) (d) 450 13000 35 Cloud Microphysics CAPE LCL 390 Cumulus Scheme 300 11000 (LNB)

LNB (LCL) 25 330 9000 200 270 7000 15 (CAPE) 100 210 5000 150 3000 5 0 1000 -200 -140 -80 -20 40 100 (km) -200 -140 -80 -20 40 100 (km) ▲ W Cyclone Centre E

Fig. 17. The zonal sections of (a) vertical velocity, ω (shade) and zonal wind (contour), (b) deviation of equivalent potential temperature from its zonal mean (taken from −200 to 100 km around the cyclone centre), (c) hourly rainfall and the contributions from the cloud microphysics and cumulus schemes, and (d) convective available potential energy (CAPE, left scale), lifting condensation level (LCL, right inner scale) and level of neutral buoyancy (LNB, right outer scale). All sections are averaged meridionally between 10 to 100 km away from the cyclone Figure 17. The zonal sections of (a) vertical velocity, ω (shade) and zonal w−i1nd (contour), (b) centre shown as thedev rediatio boxn of in eq Fig.uiva 16dlent inpo thetent matureial temp stage.erature The fro contourm its zon intervalsal mean are(tak 1.5en mfro sm -2in00(a) kmand to 0.5100 K in (b) and solid (dashed) lines denote positive km (negative)around th values.e cyclo Thene zerocentr contourse), (c) h inou(a)rly andrain(b)fallare an markedd the c ason thicktribut blackions f lines.rom t Thehe c positiveloud (negative) displacement in the abscissa refersmicr toop distancehysics an eastd cu (west)mulus ofsc thehem cyclonees, and centre(d) con atve allcti times.ve available potential energy (CAPE, left scale), lifting condensation level (LCL, right inner scale), and level of neutral buoyancy (LNB, right outer scale). All sections are averaged north-southward between 10 to 100km away from the strong conditional the cycl instabilityone centre s aroundhown asthe the centrered box compared in Fig. 16d in thtail.e ma Suchture sta g mechanisme. The contou ofr in rainterva formationls are is similar to that in to the far western 1.5 sector, m/s in especially (a) and 0.5 for K thein (b explicitly) and solid resolved (dashed) linethes den Meiyu-Baiuote positive ( frontnegativ overe) va thelues western. The North Pacific Ocean zero contours in (a) and (b) are marked as thick black lines. The positive (negative) displacement convection. Thus, it is not surprising that the microphysics (e.g. Moteki et al., 2004a, b), although the spatio-temporal in the abscissa refers to distance east (west) of the cyclone centre at all times. rain formation is overwhelmingly intense at the front. scale and the underlying frontal dynamics are very different. Regarding the comma-tail rainfall, Fig. 18 analyses the The thermodynamics indicators LNB, CAPE and LCL show dynamic and thermodynamic profiles across the meridional similar patterns to those in Fig. 17d and have extrema over confluence line in the mature stage. It shows similar fea- the confluence front. tures to those in the northeastern sector: the confluence of The diagnostics in Figs. 17 and 18 actually reflect the same southerly and northerly wind is evident in the lower tropo- underlying three-dimensional structure of the Borneo vortex sphere (Fig. 18a). Moreover, higher (lower) θe associated where the comma-shaped rainband is basically caused by the with warmer and more humid (cooler and drier) conditions is confluence of the background monsoonal cold surge and the found south (north) of the confluence line (Fig. 18b). In fact, meso-α cyclonic flow everywhere in the northern sector from the frontal nature of the confluence line is clearly marked by the comma head to the comma tail. The model simulation has a narrow zone of sharp θe gradient, indicating that southerly revealed the mesoscale details of how the moist convection (northerly) air mass originates from the equatorial South may be maintained in some Borneo vortices. The relatively China Sea (continental Asia); 70–80 % of the total rainfall at cold and dry monsoon surge wraps itself around the cyclonic the confluence front is contributed by the cloud microphysics core in the boundary layer, slowly moistening and warming scheme in the model (Fig. 18c). The peak rainfall from the up due to the surface latent and sensible heat fluxes. The air cumulus parameterization is less pronounced at the conflu- gradually loses its thermodynamic signature and transforms ence front. Similar to the comma-head rainfall, the strong up- into the warm and moist equatorial air circulating around the ward motion due to the low-level frontal confluence induces cyclone with the South China Sea as the source of moist static most of the condensation and rain formation in the comma energy. When the warm cyclonic air meets with the incoming www.atmos-chem-phys.net/14/4539/2014/ Atmos. Chem. Phys., 14, 4539–4562, 2014 4556 S. Koseki et al.: Borneo vortex and mesoscale convective rainfall

(hPa) (a) (b)

3.0 200 6.0 200

9.0

400 400

12.0 0.8

600 600

-0.4 800 800

0.4 -9.0 0.6 1000 1000 -150 -90 -30 30 90 150 (km) -150 -90 -30 30 90 150 (km)

-0.05 -0.04 -0.03 -0.02 -0.01 0 (hPa/hour)

(J/kg) (m) (m) (mm/hour) (c) (d) 400 360 18 Cloud Microphysics 9000 320 (LCL)

300 (NBL) 14 Cumulus Scheme 7000 10 200 280

(CAPE) CAPE 5000 6 100 LCL 240 NBL 2 3000 0 200 -150 -90 -30 30 90 150 (km) -150 -90 -30 30 90 150 (km) ▲ S Confluence Line N

Fig. 18. As in Fig. 17, but for the meridional section. All sections are taken across and averaged along the confluence lines (zero meridional lines in Fig. 16e) for x between 100 and 300 km in the mature stage and then temporally averaged. The convention of contours are also the same as in Fig. 17, except for 0.1 K for (b). The positive (negative) displacement in the abscissa refers to distance north (south) of the confluence line at all time.

1 Figure 18. As in Figs. 17, but for the meridional section. All sections are taken across and cold surge in2 theav northerneraged alo sectorng the c ofonf theluen Borneoce lines ( vortex,zero me itridionatendencyl lines in F (Fig.ig. 16 19g)e) fo comesr x betw mainlyeen 100 from and SELF (Fig. 19f). But is lifted up in3 a frontal300 km collision.in the matu Withre stag thee an helpd the ofn te loweredmporally averagain,aged. T thehe c negative-definiteonvention of contou STRNrs are a whichlso maximizes along LCL and enhanced4 sam CAPE,e as in F muchigs.17, condensationexcept for 0.1 K and for convec- (b). The posittheive ( frontnegati (Fig.ve) di 19e)splace isme likelynt in th toe a bebsc theissa origin of the conver- 5 refers to distance north (south) of the confluence line at all time. tive precipitation at the front gives rise to the formation of gence tendency. Therefore, the convergence along the con- a comma-shaped rainband in the northern sector. (The sharp fluence front (Fig. 19h) is forced originally by the devia- front separating the monsoon surge and the cyclonic air is not toric strain inherent in the confluence of the northeasterly observed in the SS composite of the reanalysis data partly surge and the southwesterly cyclonic wind in the Borneo because the low resolution of the global model used to as- vortex, and this is subsequently intensified by nonlinear self- similate the data cannot capture mesoscale frontal dynamics, enhancement dynamics. partly because closed-circulation vortices do not always form during SS and when they do, they are of different size, shape and at different locations.) 7 Concluding remarks Now, what maintains the meso-β-scale convergence and hence the rainband at the confluence front? The divergence We have investigated the Borneo vortex and mesoscale con- tendency diagnostic in Eq. (2) for the confluence front along vective rainband associated with the monsoon cold surge the comma tail is shown in Fig. 19. CENT is not significant over the equatorial South China Sea. implying that local rotation is not an important dynamical The composite analysis based on Cold Surge Index re- factor at the meso-β scale (Fig. 19a). The contribution of vealed that the strong cold surge transports absolute vortic- LAP is relatively weak and its distribution is similar to that of ity and water vapour from the higher tropical latitudes to the the much stronger VMOM (Fig. 19b and d). The latter is be- equatorial region. The daily rainfall over the South China Sea cause of the alignment between the pressure-gradient force is enhanced significantly when the strong cold surge occurs and the horizontal stress. An approximate balance between and its distribution matches roughly with vorticity and water HADD (Fig. 19c) and (VMOM+LAP) holds at the meso- vapour flux convergence in the lower troposphere. The diur- β scale. Along the confluence front, the net convergence nal cycle of rainfall showed that rainfall over the South China Sea is intensified over the whole day, whereas rainfall over

Atmos. Chem. Phys., 14, 4539–4562, 2014 www.atmos-chem-phys.net/14/4539/2014/ S. Koseki et al.: Borneo vortex and mesoscale convective rainfall 4557

(km) (a) (b) 220 CENT 220 LAP

200 200

180 180

160 160

140 140

120 120

100 100 10 70 130 190 250 (km) 10 70 130 190 250 (c) (d) 220 HADD 220 VMOM

200 200

180 180

160 160

140 140

120 120

100 100 10 70 130 190 250 10 70 130 190 250 (e) (f) 220 STRN 220 SELF

200 200

180 180

160 160

140 140

120 120

100 100 10 70 130 190 250 10 70 130 190 250 (g) (h) 220 (a)+(b)+(c)+(d)+(e)+(f) 220

200 200

180 180

160 160

140 140

120 120

100 100 10 70 130 190 250 10 70 130 190 250

-1 -9 -7 -5 -3 -1 1 ( 10-e4 s )

Fig. 19. Snapshot of the divergence tendency budget (see Eq. 2) along the meridional confluence front in the northeastern part of the meso-α cyclone for (a) CENT,1 (b)FLAP,igure (c)19.HADD, The sna(d)pshVMOM,ot of the (e)divSTRN,ergence (f)tenSELFdency andbudg(g)et (netcf. forcingequation at(2 850)) al hPaong atthe 140 h after the initialization; 2 meridional confluence front in northeastern part of the meso-α cyclone for (a) CENT, (b) LAP, (h) is divergence at same3 level(c) HA andDD time., (d) V TheMOM contouring, (e) STRN, convention (f) SELF, an isd ( theg) n sameet forc asing in at Fig.850 h 15.Pa a Thet 140 thick hours black after t contourhe is the zero meridional wind line. This region4is blackinitial boxizatio inn. Fig.(h) is 16d. divergence at same level and time. The contouring convention is the same as 5 in Fig. 15. The thick black contour is the zero meridional wind line. This region is the black box 6 in Fig. 16d. the Java Sea is reinforced only at night–morning time. This vorticity tendency revealed that vortex stretching due to is consistent with the fact that the rainfall over the Java Sea intense low-level convergence is mostly responsible for is mainly due to the interaction of the seaward land breeze growth/maintenance of the cyclone. The divergence tendency and the landward monsoon wind whereas the rainfall over the budget analysis suggested that the strong convergence around equatorial South China Sea is organized by larger-scale dy- the meso-α cyclone core is caused by the deviatoric strain namics. Other composite analyses of the Borneo vortex show and maintained by self-enhancement. intense rainfall and convergence in the north and northeast A comma-shaped rainband is seen sweeping northwest sectors of the Borneo vortex but shed little light on mesoscale and northeast from the centre of the meso-α cyclone with structure or processes. clusters of meso-β-scale rainfall cells. In the northwestern The semi-idealized experiment using NHM has shown sector, the confluence of the intense easterly due to the well-organized features of the Borneo vortex and the cyclone itself and weak easterly or westerly correspond- comma-shaped rainband over the equatorial South China ing to the background cold surge generates strong low-level Sea associated with cold surges. Diagnostics of absolute convergence and intense rainfall at the comma head. This www.atmos-chem-phys.net/14/4539/2014/ Atmos. Chem. Phys., 14, 4539–4562, 2014 4558 S. Koseki et al.: Borneo vortex and mesoscale convective rainfall convergence is responsible for the growth/maintenance of Chambers, C. R. S. and Li, T.: Simulation of formation of a near- the meso-α cyclone system shown by dynamical diagnoses. equatorial typhoon Vamei (2001), Meteorol. Atmos. Phys., 98, In the northeastern sector, cyclonic southeasterly flow (rela- 67–80, doi:10.1007/s00703-006-0229-0, 2007. tively warm and wet) collides with the northeasterly surge Chang, C.-P., Liu, C.-H., and Kuo, H.-C.: Typhoon Vamei: An equa- (relatively cool and dry) in the northeastern sector of the torial tropical cyclone Formation, Geophys. Res. Lett., 30, 1150, meso-α cyclone. The frontal confluence generates strong ris- doi:10.1029/2002GL016365, 2003. Chang, C.-P., Harr, P. A., and Chen, H.-J.: Synoptic disturbances ing motion which is associated with much condensation and over the equatorial South China Sea and western Maritime Con- rain formation in the comma tail of the rainband. Thermo- tinent during boreal winter, Mon. Weather Rev., 133, 489–503, dynamic factors such as convective available potential en- doi:10.1175/MWR-2868.1, 2005. ergy and lifting condensation level show consistency with Chen, G. T. J., Gerish, T. E., and Chang, C. -P.: Structure variations enhanced convective propensity at the confluence zones. The of the synoptic-scale cyclonic disturbances near Borneo during divergence tendency budget analysis along the confluence WMONEX period, Pap. Meteorol. Res., 9, 117–135, 1986. front indicates once more that the deviatoric strain inherent in Chen, J. M., Chang, C.-P., and Li, T.: Annual cycle of the confluent wind field spawns the convergence at the con- the South China sea surface temperature using the fluence front and self-enhancement dynamics greatly inten- NCEP/NCAR reanalysis, J. Meteorol. Soc. Jpn., 81, 879– sifies the convergence maintaining the rainband. 884, doi:10.2151/jmsj.81.879, 2003. Our results in this paper reveal new detailed features of Chen, T.-C., Tsay, J.-D., Yen, M.-C., and Matsumoto, J.: Interannual variation of the late fall in central Vietnam, J. Climate, 25, 392– the meso-β-scale structure of the Borneo vortex in the or- 413, doi:10.1175/JCLI-D-11-00068.1, 2012. ganization of the comma-shaped rainband. The dynamical Estoque, M. A.: The sea breeze as a function of the prevailing syn- mechanisms for growth and maintenance of the vortex and optic situation, J. Atmos. Sci., 19, 244–250, doi:10.1175/1520- the confluence front are also clarified. We suggest that the 0469(1962)019<0244:TSBAAF>2.0.CO;2, 1962. comma-shaped rainband (at meso-β scale) in the meso-α cy- Frank, W. M.: The structure and energetics of the tropical cy- clone is generated by the confluence of the background mon- clone I: Storm structure, Mon. Weather Rev., 105, 1119–1135, soonal cold surges (at synoptic scale) and the cyclonic flow doi:10.1175/1520-0493(1977)105<1119:TSAEOT>2.0.CO;2, (at meso-α scale). 1977. As a possible future investigation, mesoscale observations Hadi, T. W., Horinouchi, T., Tsuda, T., Hashiguchi, H., and of Borneo vortices would help to elucidate how realistic our Fukao. S.: Sea-breeze circulation over Jakarta, Indonesia: simulation results are. Moreover, because the resolution of A climatology based on boundary layer radar observa- tions. Mon. Weather Rev., 130, 2153–2166, doi:10.1175/1520- our experiment is only 10 km, we have not explored the 0493(2002)130<2153:SBCOJI>2.0.CO;2, 2002. meso-γ -scale (1–10 km) dynamics in the convective towers Hattori, M., Mori, S., and Matsumoto, J.: The Cross-Equatorial of the rainband in the Borneo vortex. Advanced numerical Northerly Surge over the Maritime Continent and Its Relation- experiment with finer resolution would seem desirable. ship to Precipitation Patterns, J. Meteorol. Soc. Jpn., 89A, 27–47, doi:10.2151/jmsj.2011-A02, 2011. Hayashi, S., Aranami, K., and Saito, K.: Statistical Verification of Acknowledgements. The authors would like to show their appreci- Short Term NWP by NHM and WRF-ARW with 20 km Hori- ation to two anonymous reviewers for their constructive reviewing zontal Resolution around Japan and Southeast Asia, SOLA, 4, comments and suggestions. The authors are grateful to K. Saito 133–136, doi:10.2151/sola.2008-034, 2008. and S. Hayashi of the Meteorological Research Institute, Japan Holland, G. J. and Merrill, R. T.: On the dynamics of tropical cy- Meteorological Agency, M. Sawada of Tohoku University, and clone structural-changes, Q. J. Roy. Meteorol. Soc., 110, 723– M. Inatsu of Hokkaido University, for their advice in running 745, doi:10.1002/qj.49711046510, 1984. NHM. The authors would also like to express appreciation to Hong, C.-C., Hsu, H.-H., and Chia, H.-H.: A Study of K. K. W. Cheung of Macquarie University and S. Otsuka of Kyoto East Asian Cold Surges during the 2004/05 Winter: Im- University for meaningful discussions about the Borneo vortex. pact of East Asian Jet Stream and Subtropical Upper-Level This work comprises in part Earth Observatory of Singapore Rossby Wave Trains, Terr. Atmos. Ocean. Sci., 20, 333–343, contribution 69. doi:10.3319/TAO.2008.02.04.01(A), 2009. Hubert, L. F., Kruger, A. F., and Winston, J. S.: Dou- Edited by: P. Haynes ble Intertropical Convergence Zone-Fact and Fic- tion?, J. Atmos. Sci., 26, 771–773, doi:10.1175/1520- 0469(1969)026<0771:TDICZF>2.0.CO;2, 1969. References Huffman, G. J., Adler, R. F., Bolvin, D. T., Gu, G. J., Nelkin, E. J., Bowman, K. P., Hong, Y., Stocker, E. F., and Wolff, D. B.: Asai, T.: Meso-scale features of heavy snowfalls in Japan Sea The TRMM multisatellite precipitation analysis (TMPA): Quasi- coastal region of Japan, Tenki, 35, 156–161, 1988. global, multiyear, combined-sensor precipitation estimates at Braesicke, P., Ooi, S. H., and Samah, A. A.: Propeties of strong off- fine scales, J. Hydrometeorol., 8, 38–55, doi:10.1175/JHM560.1, shore Borneo vortices: a composite analysis of flow pattern and 2007. composition as captured by ERA-Interim, Atmos. Sci. Lett., 13, 128–132, doi:10.1002/asl.372, 2012.

Atmos. Chem. Phys., 14, 4539–4562, 2014 www.atmos-chem-phys.net/14/4539/2014/ S. Koseki et al.: Borneo vortex and mesoscale convective rainfall 4559

Ikawa, M. and Saito, K.: Description of a non-hydrostatic model Nagata, M.: On the structure of a convergent cloud band over the developed at the Forecast Research Department of the MRI, Japan Sea in winter: A prediction experiment, J. Meteorol. Soc. MRI Tech. Rep. 28, Meteorological Research Institute, Tsukuba, Jpn., 65, 871–883, 1987. Ibaraki, Japan, p. 238, 1991. Nagata, M.: Meso-beta-scale vortices developing along the Japan- Johnson, R. H. and Houze Jr., R. A.: Precipitating cloud systems of Sea polar-airmass convergence zone (JPCZ) cloud band: Numer- the Asian monsoon, in: Monsoon Meteorology, edited by: Chang, ical simulation, J. Meteorol. Soc. Jpn., 71, 43–57, 1993. C.-P. and Krishnamurti, T. N., Oxford University Press, 298–353, Nakanishi, M.: Improvement of the Mellor-Yamada turbulence clo- 1987. sure model based onlarge-eddy simulation data, Bound.-Lay. Joseph, B., Bhatt, B. C., Koh, T.-Y., and Chen, S.: Sea breeze sim- Meteorol., 99, 349–378, doi:10.1023/A:1018915827400, 2001. ulation over the Malay Peninsula in an intermonsoon period, J. Nakanishi, M. and Niino, H.: An improved Mellor-Yamada Geophys. Res., 113, D20122, doi:10.1029/2008JD010319, 2008. level-3 model with condensation physics: Its design Joung, C. H. and Hitchman, M. H.: On the role of succes- and verification, Bound.-Lay. Meteorol., 112, 1–31, sive downstream development in East Asian polar air out- doi:10.1023/B:BOUN.0000020164.04146.98, 2004. breaks, Mon. Weather Rev., 110, 1224–1237, doi:10.1175/1520- Nakanishi, M. and Niino, H.: An improved Mellor-Yamada level-3 0493(1982)110<1224:OTROSD>2.0.CO;2, 1982. model: Its numerical stability and application to a regional pre- Juneng, L. and Tangang, F. T.: Long-term trends of winter monsoon diction of advection fog, Bound.-Lay. Meteorol., 119, 397–407, synoptic circulations over the maritime continent: 1962–2007, doi:10.1007/s10546-005-9030-8, 2006. Atmos. Sci. Lett., 11, 199–203, doi:10.1002/asl.272, 2010. Onogi, K., Tslttsui, J., Koide, H., Sakamoto, M., Kobayashi, S., Hat- Kain, J. S.: The Kain-Fritsch convective parameterization: An sushika, H., Matsumoto, T., Yamazaki, N., Kamahori, H., Taka- update, J. Appl. Meteorol., 43, 170–181, doi:10.1175/1520- hashi, K., Kadokura, S., Wada, K., Kato, K., Oyama, R., Ose, T., 0450(2004)043<0170:TKCPAU>2.0.CO;2, 2004. Mannoji, N., and Taira, R.: The JRA-25 Reanalysis, J. Meteorol. Kawashima, M. and Fujiyoshi, Y.: Shear instability wave along Soc. Jpn., 85, 369–432, doi:10.2151/jmsj.85.369, 2007. a snowband: Instability structure, evolution, and energetic de- Ooyama, K.: A dynamical model for the study of tropical cyclone rived from dual-Doppler radar data, J. Atmos. Sci., 62, 351–370, development, Geofisica Int., 4, 187–198, 1964. doi:10.1175/JAS-3392.1, 2005. Qian, J. H., Robertson, A. W., and Moron, V.: Interactions among Koseki, S., Koh, T.-Y., and Teo, C.-K.: Effects of the cold tongue in ENSO, the Monsoon, and Diurnal Cycle in Rainfall Vari- the South China Sea on the monsoon, diurnal cycle and rainfall ability over Java, Indonesia, J. Atmos. Sci., 67, 3509–3524, in the Maritime Continent, Q. J. Roy. Meteorol. Soc., 139, 1566– doi:10.1175/2010JAS3348.1, 2010. 1582, doi:10.1029/qj2025, 2013. Saito, K., Fujita, T., Yamada, Y., Ishida, J. I., Kumagai, Y., Aranami, Kumagai, Y.: Improvement of the land surface processes in K., Ohmori, S., Nagasawa, R., Kumagai, S., Muroi, C., Kato, JMANHM, CAS/JSC WGNE Res. Activ. Atmos. Ocean. Mod- T., Eito, H., and Yamazaki, Y.: The operational JMA nonhydro- ell., 34, 0419–0420, 2004. static mesoscale model, Mon. Weather Rev., 134, 1266–1298, Lau, K.-M. and Chang, C.-P.: Planetary scale aspects of winter mon- doi:10.1175/MWR3120.1, 2006. soon and Teleconnections, in: Monsoon Meteorology, edited by: Saito, K., Ishida, J. I., Aranami, K., Hara, T., Segawa, T., Narita, Chang, C.-P. and Krishnamurti, T. N., Oxford University Press, M., and Honda, Y.: Nonhydrostatic atmospheric models and op- 161–202, 1987. erational development at JMA, J. Meteorol. Soc. Jpn., 85B, 271– Lim, H. and Chang, C.-P.: A Theory of Mid-latitude 304, doi:10.2151/jmsj.85B.271, 2007. Forcing of Tropical Motions during Winter Mon- Seko, H., Hayashi, S., Kunii, M., and Saito, K.: Structure of the Re- soon, J. Atmos. Sci., 38, 2377–2392, doi:10.1175/1520- gional Heavy Rainfall System that Occurred in Mumbai, India, 0469(1981)038<2377:ATFMFO>2.0.CO;2, 1981. on 26 July 2005, SOLA, 4, 129–132, doi:10.2151/sola.2008-033, Lu, M.-M., and Chang, C.-P.: Unusual Late-Season Asian Cold 2008. Surges during the 2005 Asian Winter Monsoon: Roles of At- Takaya, K. and Nakamura, H.: Mechanisms of intraseasonal ampli- lantic Blocking and the Central Asian Anticyclone, J. Climate, fication of the cold Siberian high, J. Atmos. Sci., 62, 4423–4440, 22, 5205–5217, doi:10.1175/2009JCLI2935.1, 2009. doi:10.1175/JAS3629.1, 2005. Mak, M. K. and Walsh, J. E.: On the Relative Intensities of Sea and Tangang, F. T., Juneng, L., Salimun, E., Vinayachandran, P. N., Land Breezes, J. Atmos. Sci., 33, 242–251, doi:10.1175/1520- Seng, Y. K., Reason, C. J. C., Behera, S. K., and Yasunari, T.: 0469(1976)033<0242:OTRIOS>2.0.CO;2, 1976. On the roles of the northeast cold surge, the Borneo vortex, the Moteki, Q, Uyeda, H., Maesaka, T., Shinoda, T., Yoshizaki, M., and Madden-Julian Oscillation, and the Indian Ocean Dipole during Kato, T.: Structure and development of two merged rainbands the extreme 2006/2007 flood in southern Peninsular Malaysia, observed over the East China Sea during X-BAIU-99 part I: Geophys. Res. Lett., 35, L14S07, doi:10.1029/2008GL033429, Meso-beta-scale structure and development processes, J. Mete- 2008. orol. Soc. Jpn., 82, 19–44, doi:10.2151/jmsj.82.19, 2004a. Teo, C.-K., Koh, T.-Y., Lo, J. C. F., and Bhatt, B. C.: Princi- Moteki, Q, Uyeda, H., Maesaka, T., Shinoda, T., Yoshizaki, M., and pal component analysis of observed and modeled diurnal rain- Kato, T.: Structure and development of two merged rainbands fall in the Maritime Continent, J. Climate, 24, 4662–4675, observed over the East China Sea during X-BAIU-99 part II: doi:10.1175/2011JCLI4047.1, 2011. Meso-alpha-scale structure and build-up processes of conver- Tomita, T. and Yasunari, T.: Role of the northeast winter monsoon gence in the Baiu frontal region, J. Meteorol. Soc. Jpn., 82, 45– on the biennial oscillation of the ENSO/monsoon system, J. Me- 65, doi:10.2151/jmsj.82.45, 2004b. teorol. Soc. Jpn., 74, 399–413, 1996.

www.atmos-chem-phys.net/14/4539/2014/ Atmos. Chem. Phys., 14, 4539–4562, 2014 4560 S. Koseki et al.: Borneo vortex and mesoscale convective rainfall

Trilaksono, N. J., Otsuka, S., and Yoden, S.: A Time-Lagged En- Wangwongchai, A., Zhao, S. X., and Zeng, Q. C.: A case study on a semble Simulation on the Modulation of Precipitation over West strong tropical disturbance and record heavy rainfall in Hat Yai, Java in January–February 2007, Mon. Weather Rev., 140, 601– Thailand during the winter monsoon, Adv. Atmos. Sci., 22, 436– 616, doi:10.1175/MWR-D-11-00094.1, 2012. 450, doi:10.1007/BF02918757, 2005. Tsuboki, K. and Asai, T.: The multi-scale structure and de- Yamazaki, M.: A future study of the tropical cyclone without pa- velopment mechanism of mesoscale cyclones over the sea rameterizing the effects of cumulus convection, Papers Meteor. of Japan in winter, J. Meteorol. Soc. Jpn., 82, 597–621, Geophys., 34, 221–260, 1983. doi:10.2151/jmsj.2004.597, 2004.

Atmos. Chem. Phys., 14, 4539–4562, 2014 www.atmos-chem-phys.net/14/4539/2014/ S. Koseki et al.: Borneo vortex and mesoscale convective rainfall 4561

Appendix A where ui,j, and vi,j are average zonal and meridional winds from 800 to 950 hPa. Here, i and j are the zonal and merid- Definition of the centre of the meso-α cyclone ional indices of a grid point, respectively. The magnitude of the mean wind is thus The centre of the Borneo vortex and meso-α cyclone in q Sects. 4 and 5 is defined in this section. First, the mean low- = 2 + 2 Ui,j Ui,j Vi,j . (A2) level wind (Ui,j , Vi,j ) over a centred 3 × 3-point moving gird is defined as We define the location of the minimum Ui,j in the cyclone as 1 1 the cyclone centre at every hour, because the cyclonic flow 1 X X
in the immediate neighbourhood of the centre nearly cancels Ui,j = ui+k,j+l , 9 k=−1 l=−1 out, leaving sometimes a weak background wind. 1 1 1 X X
Vi,j = vi+k,j+l , (A1) 9 k=−1 l=−1

www.atmos-chem-phys.net/14/4539/2014/ Atmos. Chem. Phys., 14, 4539–4562, 2014 4562 S. Koseki et al.: Borneo vortex and mesoscale convective rainfall

Appendix B where ζ ≡ vx − uy is the relative vorticity on an isobaric sur- face. Thus, the tendency equation for absolute vorticity on an Deviation of the tendency equations for absolute vorticity isobaric surface ζa ≡ f + ζ can be derived as in Eq. (1). and divergence Taking the sum of the isobaric derivatives ∂(Eq. B1)/∂x and ∂(Eq. B2)/∂y gives The horizontal momentum equations in the frame co-moving ∂D ∂D ∂ (V −V ) with the cyclone centre are =−(V −V )·∇ D−ω −∇ ω· H C ∂t H C p ∂p p ∂p ∂u ∂u ∂u ∂u +(u−uc) +(v−vc) +ω =f (v−vc) 2 1 2 2 ∂t ∂x ∂y ∂p −D +2J (u,v)+f ζ+ f −∇ 8−β (u−uc), (B3) 2 p 2 ∂8 + 1x− −Ax, (B1) ≡ + z ∂x where D ∂u/∂x ∂v/∂y is the isobaric divergence and ∂v ∂v ∂v ∂v J (u, v) ≡ ux vy − vx uy is the Jacobian of the horizontal +(u−u ) +(v−v ) +ω =−f (u−u ) ∂t c ∂x c ∂y ∂p c wind. Now, the Jacobian can be rewritten as ∂8 2 1  2 2 2 +z1y− −Ay, (B2) J (u,v)= D +ζ −λ , (B4) ∂y 4 where for convenience, beta-plane dynamics is assumed. where λ is the deviatoric horizontal strain rate, i.e. λ and −λ Cartesian coordinates (x, y) ≡ (x0 + 1x, y0 + 1y) such that are eigenvalues of the deviatoric strain tensor S: 1x  x0, 1y  y0 represent displacements in the east–west  a b  and north–south directions respectively. u and v are zonal S≡ , and meridional winds respectively; ω ≡ Dp/Dt is the cross- b −a isobaric velocity in p coordinate where p denotes pres- where a ≡ 1 (u − v ), b ≡ 1 (v + u ). S measures the de- sure. 8 is the geopotential. f ≡ f (y0) + β 1y is the latitude- 2 x y 2 x y formation of the flow field less the effect of horizontal diver- dependent Coriolis parameter, where f0 and β take the con- gence. Moreover, by the continuity equation, −∂ω/∂p = D, stant values characteristic of the reference latitude at y0. 2 2 the second and third terms in the right-hand side of Eq. (B3) (z 1x, z 1y) is the f plane representation of the cen- ≡ 1 can be rewritten as trifugal acceleration, where z 2 f (y0) is the Earth’s ro- tation rate in the local vertical direction at y0. (The beta- ∂D ∂ (V −V ) ∂D ∂ (V −V ) −ω −∇ ω· H C =−ω +ω∇ · H C plane representation of centrifugal acceleration is not nec- ∂p p ∂p ∂p p ∂p 2 2 essary or consistent here as O(1x , 1x 1y, 1y ) terms are  ∂ (V −V )  ∂  neglected.) V ≡ (u , v ) and A ≡ (A , A ) are the veloc- −∇ · ω H C =∇ · − [ω (V −V )] C c c C x y p ∂p p ∂p H C ity and acceleration in the motion of the cyclone centre in 2 Fig. 10l and so are functions of time t only. −(V H−V C)·∇pD−D . (B5) Taking the difference of the isobaric derivatives, ∂(Eq. B1)/∂y minus ∂(Eq. B2)/∂x, gives Combining Eqs. (B4) and (B5) and using ζa ≡ f + ζ, the ten- dency equation for horizontal divergence can be derived as in ∂ζ ∂ζ ∂ω ∂v ∂ω ∂u Eq. (2). =−(V −V )·∇ ζ−ω −ζ D− − ∂t H C p ∂p ∂x ∂p ∂y ∂p

−f D−β (v−vc),

Atmos. Chem. Phys., 14, 4539–4562, 2014 www.atmos-chem-phys.net/14/4539/2014/